HCIA-Storage V5.0 Training Material

Storage Technology Trends
Foreword
⚫ Data carries information during the transmission on networks. What is the

relationship between information and data? What is the function of data
storage? This course describes the definitions of information and data in the
computer field, their relationship, as well as the concept, development
history, and development trend of data storage.
2 Huawei Confidential
Objectives
On completion of this course, you will be able to understand:

 Definitions of information and data
 Concept of data storage
 Development history of data storage
 Development trend of data storage products
Contents
1. Data and Information
2. Data Storage
3. Development of Storage Technologies
4. Development Trend of Storage Products
What is Data
⚫ SNIA (Storage Networking Industry Association) defines data as the digital
representation of anything in any form.
Format in which data is stored
Email 0101001010100010000011
1100011100010001110001
1100000111101010100101
Digital music 0101001010100101001010
1001010101001010100010
1010010101001010101010
Digital video
0101010101010010100010
0101001010101010100101
Ebook
0101001010101010100101
0101010010100101001000
1010101001001010010010
Data Types
Semi-
Structured structured Unstructured
data data data
Data Processing Cycle
⚫ Data processing is the reorganization or reordering of data by humans or
machines to increase their specific value. A data processing cycle includes three
basic steps: input, processing, and output.
Input Processing Output
What is Information
⚫ Information is processed, structured, or
rendered in a given context to make it
Processing Information
meaningful and useful.
⚫ Information is processed data, including
data with context, relevance, and
Data
purpose. It also involves the manipulation
of raw data.
Data
Redundancy
Data vs. Information
⚫ After being processed, data can be converted into information.
⚫ In order to be stored and transmitted in IT systems, information needs to be
processed as data.
Item Data Information

Raw and meaningless, with no
Feature Valuable and logical
specific purpose
Essence Original materials Processed data
Dependence Data never depends on information Information depends on data
Meteorological data or satellite
Example Weather forecasts
image data
Information Lifecycle Management
⚫ Information lifecycle management (ILM) refers to a set of management
theories and methods from the stage in which the information is generated and
initially stored to the stage where the information is obsoletely deleted.
Data value
Data Data Data Data Data Data

creation protection access migration archiving destruction
Contents
2. Data Storage
What is Data Storage
Storage in a narrow sense Storage in a broad sense
Data Storage System
⚫ Disaster recovery (DR) solutions

⚫ Backup solutions
⚫ Storage management software

Solutions
⚫ Snapshot and mirroring software
⚫ Backup software
⚫ Multipathing software
Storage software
⚫ Storage devices ⚫ Connection elements
 Disk array  HBA cards
 Tape library  Switches
Storage hardware  Virtual tape  Cables
library  …
 …
Physical Structure of Storage
Hosts
Network switching devices NAS gateway
Storage application
software
Front-end boards of
controllers
Cache
Storage
System Back-end boards of Storage management
controllers software
Connectors
Disks
Data Storage Types
Internal DAS SAN

Storage
FAS NAS
Storage
External
Storage Object Storage
Evolution of Data Management Technologies
Scale-out
Dedicated storage device
storage device
Tape and single

disk
Big data
management
Card and paper Traditional
tape database
File system system
management management
Manual
management
Data Storage Application
Database servers
Users
Application
servers
Email servers
Offline
Online storage Nearline storage
storage
File servers
Employees
Classification based on data access frequencies
Data Data Data

generation processing management
Contents
2. Data Storage

◼ Storage Architecture
▫ Storage Media
▫ Interface Protocols
History of Storage Architecture Development
2000s
1990s
• Scale-out storage
• Storage • Cloud storage
1980s network
• External
storage
1950s
• Traditional
storage
From Disks to Disk Arrays
Disks in a server Early external storage Storage arrays

Server Server Server
CPU CPU CPU
RAM RAM RAM
Disks Disks Disks
JBOD
Controllers
From Separation to Convergence
SAN NAS Clients Unified

Storage
Server Server
Ethernet
File
system Server
SAN LAN
File
NAS SAN
system
Storage Storage
File
system
Storage
Scale-out Storage
⚫ Physical resources are organized using
software to form a high-performance
Front-end
logical storage pool, ensuring reliability service
network Management
and providing multiple storage services. network
⚫ Generally, scale-out storage scatters

data to multiple independent storage
servers in a scalable system structure. It
uses those storage servers to share
...
storage loads and uses location servers
to locate storage information. Node Node Node Node
Back-end storage network
Storage Virtualization
⚫ Storage virtualization consolidates the Clients or servers
storage devices into logical resources,

thereby providing comprehensive and
unified storage services. Management
⚫ Unified functions are provided Logical resources
regardless of different storage forms
and device types.
Virtualization
Physical resources
Cloud Storage
⚫ The cloud storage system combines multiple storage devices, applications, and services. It uses
highly virtualized multi-tenant infrastructure to provide scalable storage resources for enterprises.
Those storage resources can be dynamically configured based on organization requirements.
Legacy New
applications applications
Protocol and interface
Converged
Elastic Block storage service File storage service Object storage service
Open Storage device and storage node
Universal hardware platforms
Contents
2. Data Storage

▫ Storage Architecture
◼ Storage Media
▫ Interface Protocols
History of HDDs
⚫ Larger capacity with the smaller size.
1950s–1960s 1960s–1970s 1970s–1980s 1980s-1990s 1990s-

Giant disk 14-inch disk 8-inch disk Portable Microdrive
IBM 350 RAMAC IBM 3340 Disks started to 5.25-inch 1.8-inch
IBM 1301 enter the civil 3.5-inch Hitachi: 1-inch
market. 2.5-inch Toshiba: 0.85-
inch
...
History of SSDs
⚫ Solid-state drives (SSDs) were invented almost as early as HDDs, but were not
popular at that time due to its high price and the rapid development of HDDs
at the end of the 20th century.
⚫ With the requirement for high access speed, SSDs are booming.
• Bell Labs
1976 • SanDisk
1999 • Samsung
2010…
• Floating gate • Dataram sold launched flash • BiTMICRO released laptops • Lower costs
transistor Bulk Core. SSDs. launched flash using SSDs.
• Greater capacity
• 2 MB capacity • 20 MB capacity SSDs. • 32 GB capacity
• 18 GB capacity
1967 1991 2006
Development of Flash Memory
Single-level cell Triple-level cell Quad-level cell

(SLC): fast (TLC): mature (QLC): larger and
read/write and mainstream cheaper, applicable
long service life technology to warm data
Multi-level cell 3D TLC: improves SCM: with the access

the storage density speed slightly slower than
(MLC): moderate
through multi- memory, but much faster
read/write speed than NAND media
and service life layer overlaying
Storage Class Memory (SCM)
⚫ Storage class memory (SCM) is non-volatile memory, which is slightly slower than memory but
much faster than NAND in terms of the access speed.
⚫ There are various types of SCM media under development, but the mainstream SCM media are
PCRAM, ReRAM, MRAM, and NRAM.
Current Storage Technology SCM

DRAM NAND Flash PCRAM ReRAM MRAM NRAM
Non-volatility No Yes Yes Yes Yes Yes
Read latency 10–60 ns 25 μs 48 ns < 10 ns < 10 ns < 30 ns
Write latency 10–60 ns 200 μs 40–150 ns ~ 10 ns 12.5 ns 50 ns
Erasable times > 1015 104 108 105 > 1015 > 1014
Addressing unit Byte Page Byte Byte Byte Byte
Contents
2. Data Storage

▫ Storage Architecture
▫ Storage Media
◼ Interface Protocols
Interface Protocols
⚫ Disk interfaces connect disks to hosts.
⚫ Interface protocols refer to the communication modes and requirements that
interfaces for exchanging information must comply with.
Parallel interface Serial interface
History of Interface Protocols
100G NVMe-oF 400G

Ethernet NVMe-oF 1.0 (FC-NVMe) Ethernet
Host interface (RDMA) NVMe-oF 1.1
(TCP) 64G FC
2G FC iSCSI 4G FC 8G FC 16G FC 32G FC
The first-gen FC
protocol was 2004 2008 2012 2016 2017 2018 2020…
released in 1994.
SATA 1.0
Disk interface SATA 3.0
4G FC
SAS 1.0 SAS 2.0 SAS 3.0 SAS 4.0
SATA 2.0
NVMe 1.0 NVMe 1.2 NVMe 1.3 NVMe 1.4
NVMe and NVMe-oF
⚫ NVMe, Non-Volatile Memory Express
 Improves the performance CPU CPU
 Reduces the latency PCIe PCIe
⚫ NVMe-oF, NVMe over Fabrics
SAS Controller
 Potential: low latency and high
SAS Expander
bandwidth
 Purpose: accelerates the data
transmission among the storage NVMe back-end SAS back-end
physical path physical path
network
Contents
2. Data Storage
History of Storage Products
Cloud
Media
HDD SSD
Architecture Trend
Centralized Scale-out
All-flash Intelligence
Management
Manual O&M Intelligent O&M
The Intelligence Era is Coming
Steam Age Electricity Age Information Age Intelligence Age
Industry Cloud Service extension

computing AI
Network coordination
Agriculture
+ Data + Intelligence
Internet
IoT
Intelligent production
Service Big data Customization
Challenges to Data Storage
Time-consuming RAID
24/7 µs-level
Always-on Real-time, high-speed
reconstruction services for all processing of production
Service things connected transactions in the 5G era
interruption due
to multiple points
of failure Finance High disk
Reliable Fast
failure rate
Gov & Public Large

enterprise
Low data
deduplication and
compression ratios
Large Efficient
Carrier Transportation
Unstable latency EB-level TCO reduction

Energy & Electricity Data monetized as Edge-center-cloud
under high
assets and intelligent synergy, minimizing
Inefficient convergence of concurrency storage tiering, the per-bit data cost
block, file, and object turning cold data to and maximizing the
services warm data to mine value
data value
Characteristics of Storage in the Intelligence Era
Storage in the intelligence
era
Intelligent data lifecycle
management
Storage for AI AI in storage
Data convergence
Zero data migration Data collaboration
Convergence of Simplified protocol stack and data-
SAN and NAS storage convergence
Intelligent storage tiering of hot, Compute-storage collaboration, on-
warm, and, cold data demand capacity expansion
Multi-service convergence Dedicated hardware and
(block, file, and object) scale-out storage
Hardware + Algorithm + Architecture = Efficient storage
Robust reliability Large capacity High performance
Device-level of 99.99999% PB-level to EB-level ms-level to µs-level
System-level of 99.999% Efficient deduplication & Intelligent distributed cache
compression
Solution-level reliability + Cost-effective media, warm and High-speed protocols
Cloud-level DR cold data storage and algorithms
Minimum cost and maximum value for per-bit data
Data Storage Trend
Scale-out
storage
Database Cloud
applications Blockchain
computing Storage in AI
Mainframe
the
system Hyper- Big data intelligenc
Private software Traditional
and hardware external
converged e era
infrastructure 5G
solutions with storage (HCI) General-
fully coupled purpose
compute, storage, Brain-like
computing DNA Graph
and network Tensor computing
resources, VM Container computing computing computing
applicable only to x86
large CXL HCCS
organizations RoCE/NVMe
FC IP
SAS IB Quantum dot
PCM MRAM
HDD SSD
Biomolecule Optical storage
Optical Storage Technology
Blu-ray storage Gold nanostructured glass
Long service High 100 GB+ per Low power Long-term 10 TB per
life reliability disk consumption stable storage disk
50 years >>> 600 years
DNA Data Storage
⚫ A small number of synthetic DNA molecules can store a large amount of data,
and can freeze, dry, transport, and store data for thousands of years.
⚫ Advantages of using DNA as storage media:
 Small size
 High density
 Strong stability
⚫ Bottlenecks and limitations:
 High costs of DNA molecular synthesis
 Slow data read and search
Atomic Storage
⚫ In 1959, physicist Richard Feynman suggested that it was possible to use atoms to store
information if they could be arranged the way we wanted.
⚫ Because an atom is so small, the capacity of atomic storage will be much larger than that of
existing storage media in the same size.
⚫ With the development of science and technology, arranging the atoms the way we want has
become a reality.
⚫ Bottlenecks and limitations:
 Strict requirements on the operating environment
Quantum Storage
⚫ Now, information in electronic devices is stored and moved through the flow of electrons.
⚫ If electrons are replaced by photons, the movement of information within a computer may occur
at the speed of light.
⚫ Although the storage efficiency and service life are improved, the quantum storage is still difficult
to be widely applied at present.
⚫ Quantum storage cannot meet the following requirements:
 High storage efficiency
 Low noise
 Long service life
 Operating at room temperature
Storage Network Trend
AS-IS TO-BE
FC SAN and IP SAN Converged AI Fabric network
High network costs: The FC private network has low latency but high Reduced network costs: Open Ethernet carries high-performance, low-
costs. The IP SAN has low costs but high latency and poor performance. latency, and low-cost storage networks.
High O&M costs: IP SAN and FC SAN require dedicated O&M personnel Reduced O&M costs: No dedicated O&M skills are required, and unified
separately, and do not support cloud-and-network synergy. network management is supported for data centers.
LAN Flexibility
Eth
Flexibility
No packet loss
NIC NIC
Low latency
FC SAN FC SAN
IP SAN IP SAN
10G Eth
IP FC
16G FC SAN iNIC iNIC
SAN SAN
10G IP SAN
Low cost Low latency
Ethernet
History of Huawei Storage Products
2021
2019
2017 First high-end OceanProtect
2015 Huawei OceanStor all-scenario
storage
2013 2nd-gen high- Dorado, fastest data
2011 accelerated by AI
end storage all-flash storage protection
2008 1st-gen high- chips
2006 Launched self- solution
end storage
2002 Launched 2 - nd
developed Interop Best
Released SmartMatrix
Dived into self- gen FC storage. HSSDs, pioneer architecture Best-in-class of of Show
storage developed Founded in the all-flash the year Award
technical IP SAN Huawei industry
research Symantec Co.,
Ltd.
Start out Grow Lead
Released self-
Launched developed Scale-out Huawei
FC storage virtual storage clustered Launched T OceanStor V3
2003 NAS storage series New-gen
2007 unified storage First SAN&NAS Leader in Gartner Industry's first hybrid
2009 converged Magic Quadrant flash
full-series Huawei
2012 storage for the first time storage
2014 NVMe all-flash OceanStor
2016 storage Pacific series OceanStor
2018 6.x
2020
2022
Quiz
1. (Choose multiple options) Which of the following are data types?
A. Structured data
B. Semi-structured data
C. Unstructured data
D. Massive amounts of data
2. (Choose multiple options) Which of the following statements about storage are correct?
A. Storage refers to disks.
B. A storage system consists of storage hardware, software, and solutions.
C. Storage types include block storage, file storage, and object storage.
D. File storage is used to store data of data applications.
Quiz
3. (Choose multiple options) Which of the following are characteristics of cloud storage?
A. Convergence
B. Open
C. Elasticity
D. Scale-up
4. (Choose multiple options) Which of the following are the objectives of integrating AI into storage?
A. Simple
B. Efficient
C. High power consumption
D. Easy to use
Summary
Data and Information
Data Storage Storage Architecture

Trend of Storage
Technologies
Storage Technologies Storage Media
Storage Products Interface Protocols
Recommendations
⚫ Huawei official websites

 Enterprise business: https://e.huawei.com/en/
 Technical support: https://support.huawei.com/enterprise/en/index.html
 Online learning: https://www.huawei.com/en/learning
⚫ Popular tools
 HedEx Lite
 Network Document Tool Center
 Information Query Assistant
Thank you. 把数字世界带入每个人、每个家庭、
每个组织，构建万物互联的智能世界。
Bring digital to every person, home, and
organization for a fully connected,
intelligent world.
Copyright© 2022 Huawei Technologies Co., Ltd.

All Rights Reserved.
The information in this document may contain predictive

statements including, without limitation, statements regarding
the future financial and operating results, future product
portfolio, new technology, etc. There are a number of factors
that
could cause actual results and developments to differ materially
from those expressed or implied in the predictive statements.
Therefore, such information is provided for reference purpose
only and constitutes neither an offer nor an acceptance. Huawei
may change the information at any time without notice.
Intelligent Data Storage System
Foreword
⚫ This course describes the components of the storage system, including the
controller enclosure, disk enclosure, disks, and interface modules, the
expansion methods of the storage system, and the working principles of
the storage media and the components.
Objectives
Upon completion of this course, you will understand:

 Storage product forms
 Functions and components of controller enclosures and disk enclosures
 Working principles of HDDs and SSDs
 Concepts of scale-up and scale-out, and related cables and interface modules
Contents
1. Intelligent Data Storage System
2. Intelligent Data Storage Components
3. Storage System Expansion Methods
Storage in the Intelligent Era
⚫ As we move rapidly toward an intelligent world, data is being generated at an
unprecedented rate.
⚫ More companies have realized that the key to achieving smartization is data
infrastructure, with storage at its core.
⚫ An intelligent world calls for intelligent storage.
 Storage for AI
◼ Support companies to adopt intelligent technologies to accomplish AI training and application.
 AI in Storage
◼ Integrate AI technologies into the full-lifecycle management to improve storage management,
performance, efficiency, and stability.
Intelligent Data Storage Architecture
Storage Database Big data Edge
Intelligent data reduction
• AI-based prediction
Intelligent data reduction Intelligent data storage
• CPU and GPU intelligent reduction algorithm
• Deep • Intelligent • CPU/GPU • Intelligent • Intelligent • Intelligent
compression global acceleration tiering prefetch prediction Intelligent data storage
prediction • Intelligent prefetch, data tiering, hotspot identification,
Intelligent scheduling data caching and other technologies for optimal media
Intelligence • Container • Heterogeneous computing power • Near-data computing combination
+ Intelligent scheduling
DR SC SS • Dynamic management of heterogeneous computing
Hardware CPU NPU GPU resources, near-data computing scheduling, and quick
A M D
start of containers
Memory Fabric
Network
chip
Network
chip Memory Fabric: building a memory-centric and high-
All IP performance network
• Supporting high-performance networks with nanosecond-
level latency
Memory SSD
• Memory pooling and tiering
media media
All IP: ultimate cost reduction based on SSD media
• In-depth disk-controller collaboration, and SLC, QLC, and
OLC evolution
DRAM SCM SLC/QLC QLC/OLC
• Building a simplified network with all IP
ns μs 50 μs 200 μs
Contents

◼ Controller enclosure
▫ Disk enclosure
▫ Expansion module
▫ Disk
▫ Interface module
Storage Product Form
2 U, disk and 4 U, disk and Integrated bay

controller integration controller separation
Note: Huawei OceanStor Dorado V6 is used as the example.
Controller Enclosure
⚫ The controller enclosure uses a modular design and consists of a system subrack, controllers
(with built-in fan modules), BBUs, power modules, management modules, and interface
modules. Management
Power module module
Interface
module
System subrack
Controller
Front View of a Controller Enclosure
Icon Description
Enclosure ID indicator
Enclosure location indicator

1. Blinking blue: The controller
enclosure is being located.
2 U controller enclosure (disk and controller integration) 2. Off: The controller enclosure is
not located.
Enclosure alarm indicator
1. Steady amber: An alarm is
reported by the controller
enclosure.
2. Off: The controller enclosure is
working properly.
Power indicator/Power button

4 U controller enclosure (disk and controller separation)
Rear View of a Controller Enclosure
1 2 3 4 5
No. Description
1 Management port
2 Maintenance port
6 7 3 Serial port
4 Interface module
5 Power-BBU module
6 SAS expansion port
7 Interface module
8 Management module
9 Power module
8 9
Controller
⚫ A controller is the core component of a storage system. It processes storage services,
receives configuration management commands, saves configuration data, connects
to disks, and saves critical data to coffer disks.
Controller
enclosure
FE FE
Controller Controller
Cache Cache
CPU CPU
BE BE
Disk
enclosure
Cache data
BBU and Fan Module
BBU latch Fan latch
Fan module
Running/Alarm Running/Alarm
indicator of the BBU indicator of the fan
BBU
Front view
Coffer Disk
2.5-inch coffer disk Palm-sized NVMe coffer SSD
Alarm/Location indicator Alarm/Location indicator

Running indicator Running indicator
Latch Latch
Handle Handle
Disk Disk
Label Label
Power Module
⚫ The AC power module supplies power to the controller enclosure, allowing the
enclosure to operate normally at maximum power. Running/
Alarm
Handle indicator Latch
Power socket
Contents

▫ Controller enclosure
▪ Disk enclosure
▫ Disk
Disk Enclosure
⚫ The disk enclosure uses a modular design and consists of a system subrack,
expansion modules, power modules, and disks.
Power module
Expansion
module
System subrack
Disk drive
module
Front View of a Disk Enclosure
Icon Description
ID indicator of the disk enclosure
Location indicator of the disk enclosure

2 U 25-slot smart SAS disk enclosure 1. Blinking blue: The disk enclosure is being located.
2. Off: The disk enclosure is not located.
Alarm indicator of the disk enclosure

1. Steady yellow: An alarm is reported by the disk
enclosure.
2. Off: The disk enclosure is working properly.
2 U 36-slot smart NVMe disk enclosure Power indicator of the disk enclosure
1. Steady green: The disk enclosure is powered on.
2. Off: The disk enclosure is powered off.
Power indicator/Power button
Note: This slide shows the front views of a 2 U SAS 1. The disk enclosure is powered on and off with the
controller enclosure. The power button on the disk
disk enclosure and a 2 U smart NVMe disk
enclosure is invalid and cannot be used to power
enclosure of Huawei OceanStor Dorado V6. on or off the disk enclosure separately.
Rear View of a Disk Enclosure
4
No. Description
1 Serial port
2 Mini SAS HD expansion port
3 ID display
4 Expansion module
1 2 3 5
5 Power module
2 U smart SAS disk enclosure 6 Onboard expansion port
7 Onboard management port
8 Power module
Note: This slide shows the rear

views of the 2 U smart SAS and
6 7 8
smart NVMe disk enclosures of
Huawei OceanStor Dorado V6.
2 U smart NVMe disk enclosure
Contents

▫ Disk enclosure
▪ Expansion module
▫ Disk
Expansion Module
Expansion module of a 2 U SAS disk enclosure

Connects to a controller enclosure through expansion ports
Expansion module of a 2 U smart NVMe disk enclosure

Connects to a controller enclosure through expansion ports
CE Switch
Rear view
48 x 10GE electrical ports 4 x 60 Gbit/s optical ports
Console
Front view
USB port FAN1 FAN2 PWR1 PWR2
Two Ethernet management ports (combo)
Note: Huawei CE6800 series switches are used as an example.
Fibre Channel Switch
Management ports
(serial port and
Ethernet port) USB port 24 Fibre Channel ports Power socket
Three link
aggregation groups
Note: Huawei SNS2124 is used as an example.
Device Cables
1. Serial cable 2. Mini SAS HD 3. Mini SAS HD

electrical cable optical cable
4. 100G 6. MPO-4*DLC 7. Optical fiber

QSFP28 cable 5. 25G SFP28 cable optical fiber
Contents

▫ Disk enclosure
▪ Disk
▪ HDD
▪ SSD
Disk Type
IDE
SCSI
SATA
1.8-inch SAS Interface
2.5-inch FC type
3.5-inch Dimensions
NVMe
5.25-inch ... Enterprise-class
Application
... scenario Desktop-class
...
What are the types of disks?
HDD
Structure
SSD
...
HDD Structure
⚫ An HDD consists of platters, an actuator arm, read/write heads, a spindle, a port,
and control circuits.
Platter
Actuator arm
Spindle
Control circuit
Port
Read/write head
HDD Working Principles
Basic operation
The platter is
driven by a motor.
Platter
Landing zone
Spindle Magnetic data

R/W Head
The distance between the

The head flies head and the disk is small.
over the platter.
Data Organization on a Disk
Sector Track
Head
Cylinder
Actuator arm
Platter
Motor
Motor
Disk Capacity and Cache
⚫ Disk capacity
 Disk capacity = Number of cylinders x Number of heads x Number of sectors x 512 bytes.
The unit is MB or GB. The disk capacity is determined by the capacity of a single platter
and the number of platters.
⚫ Cache
 Because the processing speed of a CPU is much faster than that of a disk, the CPU must
wait until the disk completes a read/write operation before issuing a new command. To
solve this problem, a cache is added to the disk to improve the read/write speed.
Factors Relevant to Disk Performance
Primary factor that determines the

Rotation speed throughput in the case of sequential I/Os
Primary factor that affects the random I/O

Seek speed performance
Single platter
Indirect factor for disk performance
capacity
The least important factor for disk

Port speed performance
Average Access Time
⚫ The average access time is determined by:
 Average seek time
 Average latency time
Seek time Latency Wait time
Disc
Data Block
Seek
Tracks
Data Transfer Rate
⚫ The date transfer rate is determined by:
 Internal transfer rate
 External transfer rate/Interface transfer rate
Disk Disc
Seek
Tracks
External
Internal
Disk IOPS and Transmission Bandwidth
⚫ IOPS
 Input/Output operations per second (IOPS) is a key indicator to measure disk
performance.
 IOPS is calculated by the seek time, rotation latency, and data transmission time.
⚫ Transmission bandwidth (throughput)
 Indicates the amount of data that is successfully transmitted in a unit time, that is, the
speed at which data streams are transmitted. For example, if it takes 10s to write 10,000
files of 1 KB size, the transmission bandwidth is only 1 MB/s; if it takes 0.1s to write a 10
MB file, the transmission bandwidth is 100 MB/s.
Parallel and Serial Transmission
⚫ For example, the methods for transmitting numbers 1 to 8 are as follows:
Parallel transmission Serial transmission
1
2
3 8 7 6 5 4 3 2 1
Transmit 4 Receive Transmit Receive
end 5 end end end
6
7
8
Multiple lines are connected between two ends, Only one line is connected between two ends. Eight
and one number is transmitted on each line. numbers are sent on this line in sequence. The receive
end has all numbers after eight transmissions.
HDD Port Technology
⚫ A disk must provide a simple port for users to access its data. Generally, disks
provide the following physical ports:
IDE port
Used for the ATA
instruction system
SATA port
Disk
ports Parallel SCSI port
Used for the SCSI

Serial SCSI (SAS) port
instruction system
Fibre Channel port
IDE Disk Port
⚫ The integrated drive electronics (IDE) port is also called the parallel ATA port.
 ATA stands for Advanced Technology Attachment.
 The ATA disk is also called the IDE disk.
 The ATA port uses the parallel ATA technology.
SATA Port
⚫ SATA is short for serial ATA.
 SATA ports use serial transmission and provide a higher rate than IDE ports.
 SATA uses a point-to-point architecture and supports hot swap.
SATA Version Line Code Transfer Rate Throughput

1.0 8b/10b 1.5 Gbit/s 150 MB/s
2.0 8b/10b 3 Gbit/s 300 MB/s
Power 3.0 8b/10b 6 Gbit/s 600 MB/s
SATA port
SCSI Port
⚫ SCSI is short for Small Computer System Interface.
SAS Port
⚫ SAS is short for Serial Attached SCSI.
 SAS is a point-to-point, full-duplex, and dual-port interface.
 SAS is backward compatible with SATA.
 Rate: 600 Mbit/s per channel
 SAS features high performance, high reliability, and powerful scalability.
Fibre Channel Port
⚫ Fibre Channel disks use the Fibre Channel arbitrated loop (FC-AL).
 FC-AL is a dual-port serial storage interface based on the SCSI protocol.
 FC-AL supports full-duplex mode.
 Fibre Channel provides a universal hardware transmission platform for upper-layer protocols (SCSI
and IP). It is a serial data transmission interface that features high speed, high reliability, low
latency, and high throughput.
40-pin Male FC-SCA II Connector
Contents

▫ Disk enclosure
▪ Disk
▫ HDD
▪ SSD
SSD Overview
⚫ Compared to HDDs, SSDs have absolute advantages in terms of performance, reliability,
power consumption, and portability. SSDs have been widely used in various industries.
⚫ SSD characteristics:
 Uses NAND flash to save data, providing a faster speed than HDDs.
 Has no mechanical structure inside, so it consumes less power, dissipates less heat, and generates
less noise.
 Its service life is determined by the number of program/erase (P/E) cycles.
SSD Architecture
⚫ An SSD consists of a control unit and a storage unit (mainly flash memory chips).
 Control unit: SSD controller, host interface, and DRAM
 Storage unit: NAND flash
RAM
Flash Flash
SSD Controller ...
Pkg Pkg
Processor
Host Host NAND Flash Flash
interface flash Pkg Pkg ...
Interconnect logic Buffer interface
manager
NAND Flash
⚫ Internal storage units in NAND flash include: Page 0 Page 0
Page 1 Page 1
LUNs, planes, blocks, pages, and cells
...
...

Page P Page P
Block 0 Block 1
⚫ Operations on the NAND flash include erase,
Page 0 Page 0
program, and read.
Logical Unit 0
Page 1 Page 1
...
...
⚫ NAND flash is a non-volatile medium. A block Page P Page P
Block 2 Block 3
must be erased before new data is written to it.
...
...
A program/erase (P/E) cycle is the process of Page 0 Page 0
Page 1 Page 1
erasing a block and then writing it again.
...
...
Page P Page P
Block B Block B+1
Page Register Page Register

Plane Plane
Address 0 Address 1
SLC, MLC, TLC, and QLC
111 1111
1110
11 110 1101
1100
1 101 1011
10 100
1010
1001
1000
011 0111
01 010 0110
0101
0 001 0100
00 000
0011
0010
0001
0000
SLC-1bit MLC-2bit TLC-3bit QLC-4bit

SLC MLC TLC QLC
1. Supports 50,000 to 100,000 1. Supports about 3,000 P/E cycles. 1. Provides higher data density and 1. The capacity is further
P/E cycles, providing the 2. The speed is slower than that of supports only several hundred to improved by 33%.
best reliability. SLC. 1,000 P/E cycles. 2. The performance and life
2. The storage capacity is 3. The storage capacity is relatively 2. The reliability and performance cycle are further reduced.
small. large. are low.
3. The cost is the highest. 4. The price is relatively low. 3. Generally used in personal
devices due to the cost
advantage, but cannot meet the
requirements of enterprise
products.
Flash Chip Data Relationship
Plane Die
1478 blocks 2 planes
Cell Page Block
146688 cells 768 pages

...
... ... ...
...
... ...
Page is the minimum read/write unit, and block is the minimum program/erase unit.
Address Mapping Management
Logical block address (LBA) No. 26, XX Road, Binjiang District, Hangzhou City,
Zhejiang Province, People's Republic of China
Physical block address (PBA) 120º 12' east longitude, 30º 16' north latitude
HDD: The relationship between LBA and PBA is fixed.

⚫ Overwrite
The Flash Translation
Layer (FTL) is responsible
SSD: The relationship between LBA and PBA is not fixed. for the conversion
⚫ Non-overwrite: A block must be erased before new data is between the LBA and PBA.
written to it. New data and old data are at different
locations.
FTL
Sector 0
Sector 3 Main controller Sector 2
FTL mapping table: saved in the
Sector 1
internal SRAM/DRAM, external
Sector 2 DRAM, or NAND flash. Sector 4
Sector 3 Sector 5
Sector 0 Sector 1
OS sector (512 FTL mapping Data is stored in the

bytes). File systems operation. The main NAND physical
read/write data in the controller maps the addresses based on
unit of 512 bytes. addresses based on the mapping table.
the mapping table.
Data Write Process on an SSD (1)
⚫ The following uses eight channels as an example to demonstrate how the host writes data to the SSD.
Channel 0 block Channel 1 block Channel 2 block Channel 3 block

Writes 4 KB
of data
Writes 16 KB
of data
4 KB
SSD controller
Data Write Process on an SSD (2)
⚫ When the SSD is full, old data must be deleted to release space for new data. When a user deletes and
writes data, data in some blocks becomes invalid or aged.
SSD controller

Data Read Process on an SSD
Reads 4 KB Channel 0 block Channel 1 block Channel 2 block Channel 3 block
of data
Reads 32 KB
of data 4 KB
SSD controller
SSD Performance Advantages
SSD Performance Advantages Power consumption under 100,000 read IOPS
I/O I/O
IP/FC SAN
Seek time
Mechanical latency
2 SSDs 250 HDDs
Power (W)
4000
About 400-fold
2000
VS
HDD storage system SSD storage system SSD FC HDD
Use of SSDs in Storage Systems
⚫ Class A applications: high-concurrency applications featuring random reads and writes, such as databases
⚫ Class B applications: large files, images, and streaming media featuring sequential reads and writes
⚫ Class C applications: data backup or rarely used applications
Access frequency
A SSD media
FC/SAS disk
SATA/NL-SAS/Tape
C
Data distribution
SCM Card
⚫ Storage class memory (SCM) is a new class of non-volatile memory. It is slightly slower than
memory but much faster than NAND in terms of the access speed.
⚫ An SCM card is a cache acceleration card of the SCM media type. To use SmartCache for
OceanStor Dorado V6 all-flash storage (6.1.0 and later versions), install an SCM card on the
controller enclosure.
Contents

▫ Disk enclosure
▫ Disk
◼ Interface module
Front-End: GE Interface Modules
GE electrical 10GE electrical 40GE interface 100GE interface

interface module interface module module module
Front-End: RoCE Interface Modules
25 Gbit/s RoCE interface module 100 Gbit/s RoCE interface module
Front-End: SmartIO Interface Modules
Back-End: 100 Gbit/s RDMA Interface Module and 12 Gbit/s SAS
Expansion Module
100 Gbit/s RDMA interface module 12 Gbit/s SAS expansion module
Scale-out: 100 Gbit/s RDMA Interface Module and 25 Gbit/s RDMA
Interface Module
25 Gbit/s RDMA interface module 100 Gbit/s RDMA interface module
Contents
1. Intelligent Data Storage System
3. Storage System Expansion Methods
Scale-up and Scale-out
⚫ Developments in enterprise IT systems and service expansion have caused data to skyrocket. The initial
configuration of storage systems cannot meet the demands, and now capacity expansion is a top
concern for system administrators. There are two capacity expansion methods: scale-up and scale-out.
The following uses Huawei storage products as an example to describe the two methods.
Storage
Storage
Devices (enclosure and disk) controller
controller
Storage Storage
controller controller
Scale-out Switch fabric

Storage
Scale-up controller Storage
controller
Scale-up Scale-out
SAS Disk Enclosure Scale-up Networking Principles
Adding a 2 U SAS disk enclosure to an existing loop
Controller
enclosure
2 U SAS
disk enclosure 0
2 U SAS
disk enclosure 1
Smart Disk Enclosure Scale-up Networking Principles
Adding a 2 U smart disk enclosure to an existing loop
Controller
enclosure
Smart disk
enclosure 0
Smart disk
enclosure 1
Scale-out Direct-Connection Networking
Connecting to the
management network
Controller enclosure 0
(existing)
Connecting to the
management network
Controller enclosure 1
(new)
Four-controller direct connection
The above figure shows the scale-out networking of Huawei OceanStor Dorado 5000 V6 and 6000 V6.
Scale-out Switched Networking
⚫ The networking can be switching networking or switch-free
networking.
⚫ The management networks involve external and internal

management networks.
⚫ The internal management network and internal data network

share the same physical network. For details, see the orange
and blue cables in the figure. QoS is used, minimizing the
impact on the management channel and data channel.
⚫ The external management network only needs to enable the

management network of two controllers in the first controller
enclosure to connect to the user's management network. There
is no need to connect controllers in other controller enclosures
to the user's management network.
⚫ The ETH management network ports in the rear of a CE6850

switch are used to connect to the user's network.
Local Write Process
1 2
SAN
1
LUN 2 PCIe switched network
4
Engine 0 Engine 1 Engine 2 Engine 3

3 5
Non-local Write Process
1 4
SAN
1 4
LUN
2 PCIe switched network
3
6
5
Engine 0 7 Engine 1 Engine 2 Engine 3
Local Read Process
1 2
SAN
LUN 2
PCIe switched network
5
8
Engine 0 Engine 1 Engine 2 Engine 3
3 4 6 7
Non-local Read Process
1 4
SAN
1 4
2 PCIe switched network

LUN
3
7
Engine 0 10 Engine 1 Engine 2 Engine 3

5 6 8 9
Quiz
1. (Multiple-answer question) What are the types of SSDs?
A. SLC
B. MLC
C. TLC
D. QLC
2. (Multiple-answer question) Which of the following can be used to measure the

performance of an HDD?
A. Disk capacity
B. Rotation speed
C. Data transfer rate
D. Average access time
Summary
Intelligent data Controller
storage system Controller
enclosure Coffer disk
Disk BBU
enclosure
Expansion module
Intelligent storage Expansion
Intelligent data
components module Device cables
storage system
HDD
Disk
SSD
Interface
module
Storage system Scale-up
expansion methods
Scale-out
Recommendations

 Technical support: https://support.huawei.com/enterprise/
⚫ Popular tools
 HedEx Lite
intelligent world.


that
RAID Technologies
Foreword
⚫ This course introduces technologies of traditional RAID and RAID 2.0+. The
evolution of RAID technologies aims at data protection and performance
improvement.
Objectives
After completing this course, you will be able to understand:

 Common RAID levels
 Different levels of data protection provided by different RAID levels
 Working principles of RAID 2.0+
 Dynamic RAID and RAID-TP
Contents
1. Traditional RAID
2. RAID 2.0+
3. Other RAID Technologies
Background
⚫ Problems in traditional computer systems must be addressed.
Instructions
processed per
second
CPU > 1 million Disks become the system performance bottleneck.
RAM > 100,000 Limited disk slots cannot provide sufficient capacity.
Data is stored on individual disks, so data reliability is low.

< 300
Disk
What Is RAID?
⚫ Redundant Array of Independent Disks (RAID) combines multiple physical disks into one
logical disk in different ways, improving read/write performance and data security.
⚫ Implementations: hardware RAID and software RAID.
How large is
a logical disk?
Logical disk
Physical Physical Physical Physical

disk disk disk disk
Data Organization Forms
⚫ Disk striping: Space in each disk is divided into multiple strips of a specific size. Written data
is also divided into blocks based on the strip size.
⚫ Strip: A strip consists of one or more consecutive sectors in a disk, and multiple strips form a
stripe.
⚫ Stripe: A stripe consists of strips of the same location or ID on multiple disks in the same
array.
Disk 1 Disk 2 Disk 3

D6 D7 D8 Stripe 2 Depth of a stripe
D3 D4 D5 Stripe 1
D0 D1 D2 Stripe 0
Data strips in Data strips in Data strips in

a disk a disk a disk
Data Protection Techniques
⚫ Mirroring: Data copies are stored on another redundant disk.
⚫ Exclusive or (XOR)
 XOR is widely used in digital electronics and computer science.
 XOR is a logical operation that outputs true only when inputs differ (one is true, the other is false).
◼ 0 ⊕ 0 = 0, 0 ⊕ 1 = 1, 1 ⊕ 0= 1, 1 ⊕ 1 = 0
Physical disk 1 Physical disk 2 Parity disk
1 1 0
0 1 1
0 0 0
XOR for redundant backup

Common RAID Levels and Classification Criteria
⚫ RAID levels use different combinations of data organization forms and data protection
techniques.
Consists of striping with double

RAID 6 distributed parity.
Consists of striping, but no
mirroring or parity. RAID 0
Common RAID levels RAID 10 RAID 0 is implemented after RAID

Consists of data mirroring, 1 is implemented.
without parity or striping. RAID 1
RAID 50 RAID 0 is implemented after RAID 5

Consists of striping with is implemented.
dedicated parity. RAID 3
Consists of striping with RAID 5

distributed parity.
How Does RAID 0 Work?
Write data to D 2, Read data from D 2,

D 3... D 3...
Write data to D 1. Read data from D 1.
D0 Write data to D 0. Read data from D 0.
D5 D 0, D 1, D 2, D 3, D 4, D 5
D4
D3
Physical disk 1 Physical disk 2
D2 D6
D4 D5 Stripe 2
D1
D2 D3 Stripe 1
D0
D0 D1 Stripe 0
Logical disk Data strips in a disk Data strips in a disk
Stripping without error check
Just a Bunch Of Disks (JBOD)

Write data to and read data from D 2.

D2 D 0, D 1, and D 2 pass through a mirror.
D1
Physical disk 1 Physical disk 2

D0
D2 D2
D1 D1
Logical disk D0 D0
Disk array with mirroring
Write data to C.
Write data to B. Read data.
Write data to A.
C
A 0, A 1, A 2, B 0, B 1, B 2, C 0, C 1, C 2
Physical disk 1 Physical disk 2 Physical disk 3 Parity disk
A C0 C1 C2 P3
B0 B1 B2 P2
A0 A1 A2 P1
Logical disk
Note: A write penalty occurs when just a small amount of new data needs to be written to one or two disks.
Write data.
Read data.
D5
D 0, D 1, D 2, D 3, D 4, D 5
D4
D3
D2
Physical disk 1 Physical disk 2 Physical disk 3
D1
P2 D4 D5
D0
D2 P1 D3
D0 D1 P0
Logical disk
Independent disk structure with distributed parity check codes
RAID 6
⚫ RAID 6
 Requires at least N + 2 (N > 2) disks and provides extremely high data reliability and
availability.
⚫ Common RAID 6 technologies:
 RAID 6 P+Q
 RAID6 DP
How Does RAID 6 P+Q Work?
⚫ P and Q parity data is calculated. A maximum of two data blocks that are lost can be
recovered using P and Q parity data. Formulas for calculating P and Q parity data are as
follows:
 P = D 0 ⊕ D 1 ⊕ D 2...
 Q = (α * D 0) ⊕ (β * D 1) ⊕ (γ * D 2)...
Physical disk 1 Physical disk 2 Physical disk 3 Physical disk 4 Physical disk 5
P1 Q1 D0 D1 D2 Stripe 0
D3 P2 Q2 D4 D5 Stripe 1
D6 D7 P3 Q3 D8 Stripe 2
D9 D 10 D 11 P4 Q4 Stripe 3
Q5 D 12 D 13 D 14 P5 Stripe 4
How Does RAID 6 DP Work?
⚫ Double parity (DP) provides fault tolerance up to two failed drives by adding another disk in addition to the
horizontal XOR parity disk used in RAID 4 to store diagonal XOR parity data.
⚫ P 0 to P 3 in the horizontal parity disk represent the horizontal parity data for respective disks.
⚫ For example, P 0 = D 0 XOR D 1 XOR D 2 XOR D 3
⚫ DP 0 to DP 3 in the diagonal parity disk represent the diagonal parity data for respective data disks and the
horizontal parity disk.
⚫ For example, DP 0 = D 0 XOR D 5 XOR D 10 XOR D 15
Physical Physical Physical Physical Horizontal Diagonal

disk 1 disk 2 disk 3 disk 4 parity disk parity disk
D0 D1 D2 D3 P0 DP 0 Stripe 0
D4 D5 D6 D7 P1 DP 1 Stripe 1
D8 D9 D 10 D 11 P2 DP 2 Stripe 2
D 12 D 13 D 14 D 15 P3 DP 3 Stripe 3
⚫ RAID 10 consists of nested RAID 1 + RAID 0 levels and allows disks to be mirrored (RAID 1)
and then striped (RAID 0). RAID 10 is also a widely used RAID level.
User data D 0, D 1, D 2, D 3, D 4, D 5
Disk mirror Disk mirror
D4 D4 D5 D5
D2 D2 D3 D3
D0 D0 D1 D1
Physical disk 1 Physical disk 2 Physical disk 3 Physical disk 4

RAID 1 RAID 1
RAID 0
⚫ RAID 50 consists of nested RAID 5 + RAID 0 levels. RAID 0 is implemented after RAID
5 is implemented.
D 0, D 1, D 2, D 3, D 4, D 5, D 6, D 7...
D 0, D 1, D 4, D 5, D 8, D 9 D 2, D 3, D 6, D 7, D 10, D
11
P4 D8 D9 P5 D 10 D 11 Stripe 2
D4 P2 D5 D6 P3 D7 Stripe 1
D0 D1 P0 D2 D3 P1 Stripe 0
Physical Physical Physical Physical Physical Physical
disk 1 disk 2 disk 3 disk 4 disk 5 disk 6
RAID 5 RAID 5
RAID 0
Contents
1. Traditional RAID
2. RAID 2.0+
RAID Evolution
Hot
spare
Traditional RAID LUN virtualization Block virtualization
How Does RAID 2.0+ Work?
Extent
LUN Extent
Extent
... ... Hot spare

Extent
space
CKG ... ...
CK ... ... ... ... ... ...
Disk
Disk 0 Disk 1 Disk k Disk n
Reconstruction
Traditional RAID (many-to-one) RAID 2.0+ (many-to-many)
01 02 03 51 52 53 CKG 0 (RAID 5)
HDD 0 HDD 5 HDD 0 04 05 06 HDD 5 54 55 56
07 08 09 57 58 59
Hot spare
11 12 13 61 62 63
HDD 1 HDD 6 RAID 5 (4 + 1) 14 15 16 HDD 6 64 65 66
HDD 1
17 18 19 67 68 69 CKG 1 (RAID 5)
21 22 23 71 72 73
HDD 2 HDD 7
HDD 2 24 25 26 HDD 7 74 75 76
27 28 29 77 78 79
31 32 33 81 82 83
HDD 3 HDD 8
HDD 3 34 35 36 HDD 8 84 85 86 CKG 2 (RAID 5)
37 38 39 87 88 89 52 13 63 74 85
HDD 4 HDD 9 41 42 43 91 92 93
HDD 4 44 45 46 HDD 9 94 95 96 Unused CK
Hot spare disk Hot spare
47 48 49 97 98 99
block space
Logical Objects
Storage pool Chunk CKG Extent Volume LUNs detected by

consisting of the host
physical disks
Tiered
Not tiered
Not tiered
Multiple types of Space provided by each Chunks from A CKG is divided Several extents
disks are added to disk is divided into fine- different disks form into spaces of a form a volume. LUNs can be created quickly.
a storage pool. grained chunks. CKGs. smaller
granularity.
Disk Domain
⚫ A disk domain is a combination of disks (which can be all disks in the array). After the disks are
combined and reserved for hot spare capacity, it provides storage resources for the storage pool.
Physical disks Disk domain #1
Tiers
High-performance tier
Performance tier
Disk domain #2
Capacity tier
Storage Pool and Tier
⚫ A storage pool is a storage resource container. The storage resources used by application servers are all from
storage pools.
⚫ A storage tier is a collection of storage media providing the same performance level in a storage pool. Different
storage tiers manage storage media of different performance levels and provide storage space for applications that
have different performance requirements.
Storage Supported RAID Level RAID Policy

Tier Type Application
Tier Disk Type
High- Best for storage of data that is RAID 1 1D + 1D, 1D + 1D + 1D + 1D
Tier 0 performance SSD frequently accessed with high
tier performance and price. 2D + 2D or 4D + 4D, which is automatically
RAID 10
selected by a storage system
Best for storage of data that is less
Performance frequently accessed with relatively RAID 3 2D + 1P, 4D + 1P, 8D + 1P
Tier 1 SAS
tier high performance and moderate
price. RAID 5 2D + 1P, 4D + 1P, 8D + 1P
Best for storage of mass data that
RAID 50 (2D + 1P) x 2, (4D + 1P) x 2, or (8D + 1P) x 2
is infrequently accessed with low
Tier 2 Capacity tier NL-SAS
performance and price, and large
RAID 6 2D + 2P, 4D + 2P, 8D + 2P, 16D + 2P
capacity per disk.
Disk Group
⚫ A disk group (DG) is a set of disks of the same type in a disk domain. The disk type
can be SSD, SAS, or NL-SAS.
SSD
Disk
type
SAS NL-SAS
Logical Drive
⚫ A logical drive (LD) is a disk that is managed by a storage system and corresponds
to a physical disk.
LD 0 LD 1 LD 2 LD 3
Disk 0 Disk 1 Disk 2 Disk 3
Chunk
⚫ A chunk (CK) is a disk space of a specified size allocated from a storage pool. It is
the basic unit of a RAID array.
Chunk Chunk
HDD 0 HDD 1 HDD 2 HDD 3 HDD 4
CKG
⚫ A chunk group (CKG) is a logical storage unit that consists of CKs from different
disks in the same DG based on the RAID algorithm. It is the minimum unit for
allocating resources from a disk domain to a storage pool.
CKG CKG
Disk Disk
CK
DG DG
Extent
⚫ Each CKG is divided into logical storage spaces of a specific and adjustable size
called extents. Extent is the minimum unit (granularity) for migration and statistics
of hot data. It is also the minimum unit for space application and release in a
storage pool.
LUN 0 (thick)
Extent
CKG
LUN 1 (thick)
Grain
⚫ When a thin LUN is created, extents are divided into grains of a fixed size. A thin
LUN allocates storage space by grains. Logical block addresses (LBAs) in a grain are
consecutive.
LUN (thin)
Extent Grain
CKG
Volume and LUN
⚫ A volume is an internal management object in a storage system.
⚫ A LUN is a storage unit that can be directly mapped to a host for data reads and writes. A
LUN is the external embodiment of a volume.
Server
LUN
Volume
Storage
Contents
1. Traditional RAID
2. RAID 2.0+
Huawei Dynamic RAID Algorithm
Common RAID algorithm Huawei dynamic RAID algorithm
⚫ When a block in a RAID array fails, recover the data in the faulty ⚫ When a block in a RAID array fails, recover and migrate the data
block, migrate all the data in the RAID array, and then shield the in the faulty block, shield the faulty block, and reconstruct a new
RAID array. RAID array using remaining blocks.
⚫ Result: A large amount of available flash memory space is wasted. ⚫ Benefit: The flash memory space is fully and effectively used.
4. Shield and obsolete the faulty RAID array, which wastes space. 4. Reconstruct a new RAID array using remaining blocks to store data.
ch 0 ch 1 ch n-1 ch n ch P ch 0 ch 1 ch n-1 ch n ch P
PBA 0 16 17 ... 100 60 P0 PBA 0 16 17 ... 100 60 PPm
0+2
1. A block in a RAID array fails. 1. A block in a RAID array fails.
PBA 1 101 160 ... 10 11 P1 PBA 1 101 160 ... 10 11 P1
PBA m 3000 1280 ... n n+1 Pm PBA m 3000 1280 ... n n+1 Pm
2. Create a new RAID array to store the data of the RAID array where a block fails. 2. Create a new RAID array to store the data in the faulty block.
PBA m+1 P m+1 PBA m+1 P m+1

3. Recover the data in the faulty block using the RAID algorithm 3. Recover the data in the faulty block using the RAID
and migrate all the data in the RAID array. algorithm and migrate the data.
RAID-TP
⚫ RAID protection is essential to a storage system for consistently high reliability and performance. However, the
reliability of RAID protection is challenged by uncontrollable RAID array construction time due to drastic increase in
capacity.
⚫ RAID-TP achieves optimal performance, reliability, and capacity utilization.
Tolerates the simultaneous failure of three RAID-TP

disks without service interruption.
Traditional
RAID
Huawei RAID-TP: • Greatly reduces reconstruction time.

• Tolerates the simultaneous failure of three disks • Effectively copes with data protection challenges
in the era of large-capacity disks.
Quiz
1. What is the difference between a strip and a stripe?
2. Which RAID level would you recommend if a user focuses on reliability and
random write performance?
3. Is it true or false that data access will remain unaffected when any disk in a RAID
10 array fails?
Summary
RAID 0, RAID 1, RAID 3, RAID 5, RAID 6

Traditional RAID
RAID 10 and RAID 50
RAID Technologies RAID 2.0+ RAID 2.0+
Dynamic RAID
Other RAID
technologies
RAID-TP
Recommendations

 Enterprise business: http://enterprise.huawei.com/en/
 Online learning: https://learning.huawei.com/en/
⚫ Popular tools
 HedEx Lite
 Network Documentation Tool Center
intelligent world.
Copyright©2022 Huawei Technologies Co., Ltd.


portfolio, new technology, etc. There are a number of factors that
Common Storage Protocols
Foreword
⚫ A protocol is a set of conventions that both computers must comply with to

communicate. For example, the protocol may establish conventions to set
up connections or identify each other.
⚫ A protocol not only defines the language used for communication, but also
specifies the hardware, transmission medium, transmission protocol, and
interface technology. This course describes the definitions and principles of
different storage protocols.
Objectives
Upon completion of this course, you will learn:

 Common protocols used in storage systems; and
 Working principles and characteristics of these protocols
Contents
1. SAN Protocols
◼ SCSI and SAS
▫ iSCSI and FC
▫ PCIe and NVMe
▫ RDMA and RoCE
2. NAS Protocols
3. Object and HDFS Storage Protocols
SCSI Protocol
⚫ Small Computer System Interface (SCSI) is a vast
SCSI device instructions
protocol system evolved from SCSI-1 to SCSI-2 and to (SBC/SSC)
its current version, SCSI-3.
⚫ It defines a model and necessary command set for
SCSI system model
different devices to exchange information using the (SAM/SPC)
framework.
⚫ Transmission media has no bearing on the SCSI
SCSI link implementation
protocol. It can be used across various media types, (FCP/SAS/iSCSI)
including virtual media.
SCSI Evolution
SCSI-1
⚫ Launched between 1983

and 1985.
⚫ Supported synchronous
SCSI-1 and asynchronous modes.
SCSI-2
⚫ Supported up to seven 8-
bit devices.
⚫ Prevalent between 1988
⚫ Ran at up to 5 Mbit/s.
and 1994.
⚫ Used 50-pin cables at a SCSI-3
⚫ Compatible with SCSI-1.
max distance of 6 m.
⚫ Supports 16-bit
Now obsolete.
SCSI-2
⚫
bandwidth. ⚫ Standardized in 1993.
⚫ Runs up to 20 Mbit/s. ⚫ Compatible with SCSI-1 and
SCSI-2.
Is a standard system.
SCSI-3
⚫
⚫ Supports various media such

as FCP and IEEE1394.
SCSI Logical Topology
Initiator
Initiator 1 Initiator 2
SDS
Target
Target 1 Target 2 Target 3
LUN 0
LUN 2
LUN 5
LUN namespace Target device
SCSI Initiator Model
⚫ A host's SCSI system typically works in the initiator mode. The SCSI architecture on
Windows, Linux, AIX, Solaris, and BSD contains the architecture (middle layer),
device, and transport layers.
SCSI device driver layer
SCSI middle Layer
SCSI link layer
SCSI Target Model
⚫ Based on the SCSI architecture, a target is divided into
three layers: port layer, middle layer, and device layer.
Port Layer
⚫ The middle layer, the most important, manages LUN
namespaces, link ports, target devices, tasks, task sets,
Middle Layer
and sessions based on SAM/SPC specifications.
⚫ Drivers at the port layer are dynamically loaded via
registration, while those at the device layer are Device Layer
dynamically loaded.
SCSI Protocol and Storage System
⚫ The SCSI protocol is the basic protocol used for communication between hosts
and storage devices.
⚫ DAS uses the SCSI protocol to achieve interconnection between hosts and
storage devices.
SCSI bus
Host adapter Data/Address bus

SCSI ID 7
Control signal
SCSI array SCSI array
ID 0 ID 5
LUN 0 LUN 1 LUN 0

LUN 2 LUN 1
SCSI Protocol Addressing
Bus Differentiates SCSI

number buses.
Differentiates devices on
Device ID
SCSI buses.
Logical
unit
Differentiates sub-
number devices in SCSI devices.
Background of SAS
• The performance of the parallel SCSI technology has plateaued and the bandwidth cannot be improved.
• Serial bus technologies, such as Fibre Channel, InfiniBand (IB), and Ethernet, have drawbacks in storage
applications to some extent:
a) Fibre Channel applies to complex networking and long-distance scenarios but requires a high budget.
b) InfiniBand requires a high budget and complex networking.
c) iSCSI has high latency and low transfer speeds.
SCSI in serial mode: SAS
What Is SAS
⚫ Serial Attached SCSI (SAS) is the serial standard of the SCSI bus protocol.
⚫ SAS uses serial technology to achieve higher transmission rate and better scalability,
and is compatible with SATA disks.
⚫ SAS adopts the point-to-point architecture to achieve a transmission rate of up to 3
Gbit/s, 6 Gbit/s, 12 Gbit/s, or higher. The full-duplex mode is supported.
SAS Protocol Layers
SCSI application layer ATA application layer Management application layer Application layer
SSP transport layer STP transport layer SMP transport layer Transport layer
SAS port layer Port layer
SSP link layer STP link layer SMP link layer

Link layer
SAS link layer
SAS phy layer PHY layer
SAS physical layer Physical layer
Highlights of SAS
⚫ Provides the serial communication mode to allow multiple data channels to
communicate at full speed with devices.
⚫ Binds multiple narrow ports to form a wide port.
⚫ Uses expanders to expand interfaces, providing excellent scalability.
⚫ Works in full-duplex mode.
Expander
Server
SAS Expander
RAID Expander
Controller Wide Expander
Link
Expander
SAS
Read Data
RAID Expander
Commands
Controller
Scalability of SAS
⚫ SAS uses expanders to expand interfaces. One SAS domain supports up to 16,384
disk devices.
1
Expander
Expander
Expander
SAS RAID
Expander
Controller
1
Expander
Expander
128
128
Cable Connection Principles of SAS
⚫ Generally, a SAS cable has four channels, each of which supports 12 Gbit/s bandwidth.
⚫ SAS devices are connected in the form of a loop (also called a chain).
⚫ The cable bandwidth is 4 x 12 Gbit/s, which limits the number of disks that can be
included in a loop.
⚫ A maximum of 168 disks are supported in a loop. That is, a loop consists of a
maximum of seven disk enclosures with 24 disk slots each.
SAS cable connectors:
Mini SAS
High-density mini SAS
Contents
1. SAN Protocols
▫ SCSI and SAS
◼ iSCSI and FC
▫ PCIe and NVMe
▫ RDMA and RoCE
2. NAS Protocols
Emergence of iSCSI
SCSI is used to connect a small number of devices at a

limited distance.
IP-network-based SCSI: iSCSI
Introduction to iSCSI
SCSI applications (file systems and databases)
SCSI block SCSI flow Other SCSI

instructions instructions instructions
SCSI commands, responses, and data
SCSI instructions, data, and status
Receives SCSI data, generates iSCSI
iSCSI protocol data, and encapsulates the
data into iSCSI PDUs.
TCP Encapsulates iSCSI PDUs into TCP/IP

packets for transmission and reception.
IP
Converts bit streams and transmits
Ethernet them over physical links.
iSCSI protocol stack
iSCSI Initiator and Target
⚫ Initiator iSCSI
Initiator Target
 The SCSI layer generates command descriptor blocks
(CDBs) and transfers them to the iSCSI layer. SCSI SCSI
 The iSCSI layer generates iSCSI protocol data units

iSCSI iSCSI
(PDUs) and sends them to the target over an IP network.
⚫ Target TCP TCP
 The iSCSI layer receives PDUs and sends CDBs to the

IP IP
SCSI layer.
 The SCSI layer interprets CDBs and gives responses when Link Link
necessary.
iSCSI Architecture
⚫ iSCSI nodes encapsulate SCSI instructions and data into iSCSI packets and send the
packets to the TCP/IP layer, where the packets are encapsulated into IP packets to
be transmitted over an IP network.
iSCSI node iSCSI node
IP address of IP address of
iSCSI
the network the network
Target
port port
iSCSI IP network
Initiator
IP address of IP address of
iSCSI
the network the network
Target
port port
Relationships Between iSCSI and SCSI, TCP and IP
initiator target
Feature description
SCSI application SCSI application
protocol SCSI commands, responses, and data protocol
Receives SCSI data, generates iSCSI

protocol data, and encapsulates the data
iSCSI protocol into iSCSI PDUs. iSCSI protocol
Encapsulates iSCSI PDUs into TCP/IP

TCP/IP protocol TCP/IP protocol
packets for transmission and reception.
stack stack
Performs 8-bit/10-bit encoding, as well
NIC driver as transmits and receives data. NIC driver
Performs 0/1 code stream transmission.

Physical Physical
FC in Storage
⚫ Fibre Channel (FC), also referred to as the FC protocol, FC network, or FC
interconnection, delivers high performance for front-end host access on point-to-
point and switch-based networks.
⚫ FC brings the following advantages to the storage network:
 Improved scalability
 Increased transmission distance
 High security
Initiator HOST Initiator
FC Switch FC Switch
Target Target
Storage Disk enclosure Disk enclosure
FC Protocol Structure
Upper-layer protocol
SCSI-3 IP ATM
FC-4 IPI-3 SCSI-3 FC-LE

command set command set Link FC-ATM
mapping mapping encapsulation
FC-3 General equipment
FC-2 Structure protocol FC-PH FC-AL FC-AL2

FC-1 Coding/Decoding FC-PH2
Copper cables and optical cables,
FC-PH3
FC-0 Physical conversion 8-bit/10-bit encoding
FC Topology
Point-to-point FC-AL FC switching network
Most widely used technology
Only two devices can be connected. Up to 127 devices can Up to 16 million devices
(Direct connection) be connected. can be connected.
Seven Port Types of the FC Protocol
⚫ There are seven types of ports in FC networks.
FC terminal N F FC switch F N FC terminal
Fabric device FL Fabric device

FC terminal FC terminal
NL NL
Open ring device Open ring device
FC-AL device
L L L
FC terminal FC terminal FC terminal
Private ring device Private ring device Private ring device
FC Adapter
⚫ The FC host bus adapter (HBA) supports FC network applications and provides high-
bandwidth and -performance storage network solutions.
Contents
1. SAN Protocols
▫ SCSI and SAS
▫ iSCSI and FC
◼ PCIe and NVMe
▫ RDMA and RoCE
2. NAS Protocols
PCIe
⚫ PCIe is short for PCI Express, which is a high-performance and high-bandwidth serial
communication interconnection standard. It was first proposed by Intel and then
developed by the Peripheral Component Interconnect Special Interest Group (PCI-
SIG) to replace the bus-based communication architecture, such as PCI, PCI Extended
(PCI-X), and Accelerated Graphics Port (AGP).
Why PCIe?
PCIe is used to significantly improve system throughput, scalability, and flexibility at

lower production costs, which are almost impossible to achieve using the traditional
bus-based interconnection.
High-performance and high-bandwidth serial interconnection standard: PCIe
PCIe Protocol Structure
⚫ PCIe device layers comprise the physical, data link, transaction, and application
layers.
PCIe Device
 Physical layer
AP
 Data link layer
 Transaction layer
Transaction
 Application layer
Physical Data Link
Logical Sub-Block
Physical
Electrical Sub-Block
Tx Rx
NVMe
⚫ NVMe is short for Non-Volatile Memory Express.
⚫ The NVMe standard is oriented to PCIe SSDs. Direct connection from the native PCIe
channel to the CPU can avoid the latency caused by communication between the
external controller (PCH) of the SATA and SAS interface and the CPU.
⚫ PCIe is an interface form and a bus standard, and NVMe is a standard interface
protocol customized for PCIe SSDs.
NVMe Protocol Stack
App
Reduced interaction: The number of communication
interactions is reduced from 4 to 2, reducing the latency.
Block Layer
Controller SSD
Controller 1. Transfer command
SCSI Initiator NVMe
2. Ready to transfer
SAS
3. Transfer data
SAS
4. Response feedback
Target 1. NVMe write command

SAS NVMe
NVMe 2. NVMe write finished
SCSI
SAS protocol stack NVMe protocol stack The average I/O latency when NVMe is used is less than
that when SAS 3.0 is used.
NVMe Working Principle (DMA)
⚫ NVMe uses Direct Memory Access (DMA) to transfer NVMe commands.
CPU 1. The NVMe driver maps the controller

NVMe host driver
3 4 register of the storage device to the
2 host memory.
Root complex
Memory Submission Completion 2. The host controls the storage device
PCIe queue queue Controller register
and reads the device status using the
register.
Switch
3. The host writes I/O requests into the
1
DMA Mapped to Host submission queue for sending NVMe
memory commands.
NVMe controller 4. The NVMe device transfers data via
Command/Response Controller register DMA and writes completion to the host.
Other registers
NVMe supports multi-queue data transmission

SSD SSD SSD SSD with high concurrency and low latency.
Storage device
Advantages and Application of NVMe
IOPS
Server Over Fabric
End-to-end 100GE
FC NVMeoF
CloudEngine
100GE NVMe-oF
CE8800 End-to-end hardware Reduced by 50%
Latency
uninstallation
Huawei NVMe-oF protocol
OceanStor
FC NVMe-
Dorado
oF
Reduced by 66%
TCO
End-to-end DIF
Data integrity protection
NVMe SSD
FC NVMeoF
Contents
1. SAN Protocols
▫ SCSI and SAS
▫ iSCSI and FC
▫ PCIe and NVMe

◼ RDMA and RoCE
2. NAS Protocols
Introduction to NVMe over Fabrics
⚫ NVMe over Fabrics (NVMe-oF) is an extension of the NVMe protocol to Ethernet and Fibre Channel. It allows
accessing remote NVMe devices as if accessing local NVMe devices.
⚫ To address the PCIe scalability problem, NVMe released a specification in 2016, which defines two types of fabric
transports, implementing the end-to-end NVMe protocol.
 NVMe over Fabrics using RDMA
 NVMe over Fabrics using Fibre Channel (FC-NVMe)
Local NVMe Host PCIe Switch NVMe SSD
PCIe PCIe
NVMe-oF Host HBA/NIC Network HBA/NIC NVMe SSD
NVMe over Fabrics
RDMA
⚫ RDMA is short for Remote Direct Server 1 Server 2
Memory Access, which is a method
User
Application Application
of transferring data in a buffer Buffer 1. Kernel bypass Buffer
between application software on Buffer 2. Offload Buffer
Kernel
OS OS
3. Zero copy
two servers over a network. Buffer Buffer
 Low latency
CPU CPU
Hardware
RDMA
 High throughput HCA HCA
 Low CPU and OS resource occupancy NIC Buffer Buffer NIC

TCP/IP
RDMA Bearer Network
RDMA Application/ULP
software
RDMA API(Verbs)
RDMA Software Stack

Typically Hardware
IB Transport Protocol IB Transport Protocol IB Transport Protocol iWARP Protocol
UDP TCP
IB Network Layer IB Network Layer
IP IP
IB Link Layer Ethernet Link Layer Ethernet Link Layer Ethernet Link Layer
InfiniBand RoCE v1 RoCE v2 iWARP
InfiniBand Ethernet/IP Ethernet/IP Ethernet/IP

Management Management Management Management
Content defined by IBTA

Content defined by IEEE/IETF
Introduction to NVMe over RoCE
⚫ NVMe over RoCE is a type of the NVMe over Fabrics protocol based on RDMA. It has been significantly
optimized in terms of performance, cost, network management, and technology development, and is
gradually becoming an optimal application of NVMe over Fabrics.
Host Multi-Path
software
Server ⚫ Linux hosts support NVMe over Fabrics Host Driver.
NVMe Driver
Host RoCE
NIC
⚫ The host NIC supports RoCEv2, and the Linux kernel supports RoCE.
Transport Driver
Network
iLossLess
Fabric
NIC RoCE ⚫ The storage system supports both NVMe over Fabrics and RoCEv2.
Transport Driver ⚫ Currently, the mainstream storage vendors include Huawei, Pure
NVM subsystem Storage, and NetApp.
Storage
Multi-Path software
Contents
1. SAN Protocols
2. NAS Protocols
◼ File System
▫ CIFS, NFS and Cross-Protocol Access
▫ HTTP, FTP and NDMP
File System
⚫ A file system is used by a computer to manage and organize data in the form of files and
directories. It forms a tree diagram, where the leaf node is a file, the intermediate node is a
directory at each level, and the top level is the root directory.
⚫ The file service is used to provide stable and reliable file sharing functions. It is a basic
feature of the NAS storage system and supports shared file access using CIFS and NFS
clients.
/root
/A /B /C
: Contents
: File
/A/B /C/C /C/A
Contents
1. SAN Protocols
2. NAS Protocols
▫ File System
◼ CIFS, NFS and Cross-Protocol Access
▫ HTTP, FTP and NDMP
CIFS Protocol
⚫ In 1996, Microsoft renamed SMB to CIFS and added many new functions. Now, CIFS
includes SMB1, SMB2, and SMB3.
⚫ CIFS uses the client/server mode and TCP/IP and IPX/SPX basic network protocols.
The CIFS share feature is primarily used by Windows-based clients to share files in a
non-domain environment or an AD domain environment.
Named as SMB Redefined
MS-DOS CIFS Defined SMB2 SMB2.2 => SMB3.0
1988 1992 1996 2000 2007 2009 2012 2020
IBM Realized Samba Added new features

Defined SMB2.1
CIFS share in non-domain environments
Maintenance
terminal
Management
network port
CIFS share
CIFS
Windows client
Storage system
CIFS share in AD domain environments
Maintenance
terminal
User identity authentication

management
AD domain controller
Management
network port
CIFS share
CIFS
Windows client
Storage system
（AD domain client） Authentication flow
Data flow
NFS Protocol
⚫ NFS, or Network File System, is a protocol defined by the IETF and widely used in
the Linux/Unix environment.
⚫ NFS works based on the client/server architecture. A server provides clients with file
system access, whereas clients access shared file systems. The NFS feature enables
clients running a variety of operating systems to share files over a network. It
applies to a wide range of network environments, including the non-domain
environment, LDAP domain environment, and NIS domain environment.
NFS V1 NFS V2 NFS V3 NFS V4.0 NFS V4.1 NFS V4.2
1990 2000 2010 2020

NFS share in a non-domain environment
Maintenance
terminal
Management
network port
Linux or UNIX client
/mnt
/Filesystem/share NFS
Storage system
VMware client
/mnt/data
NFS share in a domain environment
Maintenance
terminal
Network group
management
Management LDAP or NIS domain server

network port
/Filesystem/share
NFS
Linux or UNIX client
/mnt
Storage system
CIFS-NFS Cross-Protocol Access
SMB service client NFS service client
AD authentication UNIX security
SMB 1/SMB 2/SMB 3 NFSv3/NFSv4
DNS/AD (IDMU) server LDAP/NIS server

SMB-NFS
access
DNS query
AD authentication NIS/LDAP query
Storage system (NAS)
Contents
1. SAN Protocols
2. NAS Protocols
▫ File System
▫ CIFS, NFS and Cross-Protocol Access

◼ HTTP, FTP and NDMP
HTTP Shared File System
⚫ The storage system supports the HTTP
shared file system. With the HTTPS
service enabled, you can share a file
system in HTTPS mode.
PC
⚫ Shared resource management is
implemented based on the WebDAV
protocol. As an HTTP extension protocol, Server
WebDAV allows clients to copy, move,
modify, lock, unlock, and search for
resources in shared directories on servers. PC
File Transfer Protocol (FTP)
⚫ FTP is a universal protocol for transferring files between remote servers and local clients over an
IP network.
⚫ It belongs to the application layer in the TCP/IP protocol suite and employs TCP ports 20 and 21
to transfer files between remote servers and local clients. Port 20 is used to transfer data, and
port 21 is used to transfer control messages. RFC 959 describes the basic FTP operations.
⚫ The storage system supports the FTP service. The FTPS server function on a device allows users to
securely access a remote device by using the FTP client.
⚫ FTP works in either of the following modes:
 Active mode (PORT): The FTP server creates a data connection request. This mode is inapplicable when
the FTP client is behind a firewall, for example, the FTP client resides on a private network.
 Passive mode (PASV): The FTP client creates a data connection request. This mode is inapplicable when
the FTP server does not allow the FTP client to connect to its high-order ports (usually, the port IDs are
larger than 1024).
Active Mode of the FTP Server
192.168.10.50 Control connection in active mode 192.168.10.200
SYN
30000
ACK + SYN 20
ACK
FTP client Data connection FTP server
192.168.10.50 Control connection in active mode 192.168.10.200

SYN
ACK + SYN 21
ACK
FTP client Control connection FTP server
Passive Mode of the FTP Server
192.168.10.50 Control connection in passive mode 192.168.10.200
SYN
ACK + SYN 21
ACK
FTP client Control connection FTP server
Control connection in passive mode

192.168.10.50 192.168.10.200
SYN
ACK+SYN 30000
ACK
Data connection FTP server
FTP client
NDMP Protocol
⚫ The NDMP protocol is designed for the data
backup system of NAS devices. It enables NAS
2-way Production Production Backup server
networking system A system N devices, without any backup client agent, to send
data directly to the connected disk devices or the
NAS storage
system Tape library backup servers on the network for backup.
⚫ There are two networking modes for NDMP:
 2-way
3-way
Production Production
 3-way
networking Backup server
system A system N
NAS storage Fiber Channel

system A Ethernet
Backup data flow
NAS storage system A
Tape library Control flow
Contents
1. SAN Protocols
2. NAS Protocols

◼ Object Storage Protocol
▫ HDFS Storage Protocol
Object Service
⚫ The object service is an object-based mass data storage service offering scalable, secure, reliable, and cost-
effective data storage capabilities.
⚫ The object service provides standard S3 APIs, which are HTTP/HTTPS-based REST APIs. Users can use object
service clients, APIs, and SDKs to easily manage and use object service data and develop various types of upper-
layer applications.
Object service client API SDK
S3 S3 S3
Namespace Namespace Namespace
Object Object Object Object ... Object Object
Storage pool
...
Node 1 Node 2 Node 3 Node N
S3 Concepts
⚫ The following figure shows the relationship among buckets, objects, accounts,
and users.
Account
Network
...
User User
Bucket Bucket Bucket
Object Object
... Object
Object Object Object
RESTful
⚫ REST
 REST is short for REpresentational State Transfer. It indicates the state transfer of resources in a certain form on
the network.
◼ Resource: resource, that is, data. For example, newsfeed and friends.
◼ Representational: a representation form, for example, an image or a video.
◼ State Transfer: status change. This is implemented through HTTP verbs.
 It uses the HTTP protocol and URI to add, delete, modify, and query resources using the client/server model.
 REST is not a specification, but an architecture for network applications. It can be regarded as a design mode
which is applied to the network application architecture.
⚫ RESTful
 An architecture complying with the REST principle is called a RESTful architecture.
 It provides a set of software design guidelines and constraints for designing software for interaction between
clients and servers. RESTful software is simpler and more hierarchical and facilitates the cache mechanism.
Contents
1. SAN Protocols
2. NAS Protocols


Object Storage Protocol
◼ HDFS Storage Protocol
HDFS Service
⚫ The HDFS service provides an HDFS decoupled storage-compute solution based on native HDFS. The solution
implements on-demand configuration of storage and compute resources, provides consistent user experience, and
helps reduce the total cost of ownership (TCO). It can coexist with the legacy coupled storage-compute
architecture.
⚫ The HDFS service provides native HDFS interfaces to interconnect with big data platforms, such as FusionInsight,
Cloudera CDH, and Hortonworks HDP, to implement big data storage and computing and provide big data analysis
services for upper-layer big data applications. Big data application
Hadoop compute cluster

FusionInsight Cloudera Hortonworks
...
Management Compute Compute Compute
node node node node
Native HDFS APIs
... Scale-out
storage cluster
Distributed File System
⚫ The distributed file system stores files on

multiple computer nodes. Thousands of
computer nodes form a computer cluster.
⚫ Currently, the computer cluster used by the
distributed file system consists of common
hardware, which greatly reduces the hardware
overhead.
⚫ The Hadoop Distributed File System (HDFS) is a
distributed file system running on universal
hardware and is designed and developed based
on the GFS paper.
HDFS Architecture
HDFS Architecture
Metadata (Name, replicas, ...):

NameNode /home/foo/data,3...
Metadata ops
Block ops
Client
Read DataNode DataNode
Replication
Blocks Blocks
Client
Rack 1 Rack 2
HDFS Communication Protocol
⚫ HDFS is a distributed file system deployed on a cluster. Therefore, a large amount of
data needs to be transmitted over the network.
 All HDFS communication protocols are based on the TCP/IP protocol.
 The client initiates a TCP connection to the NameNode through a configurable port and
uses the client protocol to interact with the NameNode.
 The NameNode and the DataNode interact with each other by using the DataNode
protocol.
 The interaction between the client and the DataNode is implemented through the Remote
Procedure Call (RPC). In design, the name node does not initiate an RPC request, but
responds to RPC requests from the client and DataNode.
Quiz
1. Which networks are included in FC topologies?

A. Fibre Channel Arbitrated Loop (FC-AL)
B. Fibre Channel point-to-point (FC-P2P)
C. Switched network
D. Fibre Channel dual-switch
2. Which NFS versions are available currently?

A. NFSv1
B. NFSv2
C. NFSv3
D. NFSv4
Quiz
3. Which of the following are file sharing protocols?
A. HTTP
B. iSCSI
C. NFS
D. CIFS
4. Which of the following operations are involved in the CIFS protocol?

A. Protocol handshake
B. Security authentication
C. Share connection
D. File operations
E. Disconnection
Summary
SCSI and SAS
iSCSI and FC
SAN protocol
PCIe and NVMe
RDMA and RoCE

Common storage
protocols File system
NAS protocol CIFS, NFS and cross-protocol access
HTTP, FTP and NDMP
Object and Object storage protocol
HDFS storage HDFS storage protocol

protocols
Recommendations
⚫ Huawei official websites:
 Enterprise service: https://e.huawei.com/en/
⚫ Popular tools
 HedEx Lite
 Network documentation tool center
 Information query assistant
intelligent world.


that
Storage Network Architecture
Foreword
⚫ With the development of host, disk, and network technologies, the storage
system architecture evolves, and the storage network architecture also
develops to meet service requirements. This course introduces the storage
network architecture.
Objectives
Upon completion of this lesson, you will be able to understand:

 Storage network architecture evolution
 Storage network technology evolution
Contents
1. DAS
2. NAS
3. SAN
4. Distributed Architecture
DAS
LAN
File server Application server Database server
Client
Data flow SCSI Data flow SCSI Data flow SCSI
Storage device Storage Storage

device device
DAS
External disk array (DAS) Smart disk array (DAS)
Server Server
CPU CPU
RAM
RAM
Disk
Disk
SCSI card
RAID SCSI card
Controller
JBOD
Challenges for DAS
Challenges Description
Limited number of ports that can be connected to a host.

Low
Scalability Limited number of addressable disks.
Limited distance.
Inconvenient
The system needs to be powered off during maintenance.
Maintenance
Insufficient Front-end ports and storage space are difficult to share.

Resource
Sharing Resource silos: For example, the DAS with insufficient storage
space cannot share the remaining space of the DAS with excessive
storage resources.
Contents
1. DAS
2. NAS
3. SAN
NAS
⚫ Network-attached storage (NAS) connects storage devices to the
live network and provides data and file services.
⚫ The most commonly used network sharing protocols for NAS are
Common Internet File System (CIFS) and Network File System
(NFS). ⚫ Benefits:
OS: Windows  Improved efficiency
 Improved flexibility
 Centralized storage
OS: Linux  Simplified management
 High scalability
 High availability
Switch NAS device
OS: MAC OS  Security (user authentication
and authorization)
General-Purpose Server and NAS Devices
File system
Application
OS
Printer driver
Network
File system
OS
Network Uni-functional
NAS device
General-purpose server
(Windows or UNIX)
NAS Protocols
Enterprise Enterprise
office office • NFS
• CIFS
IP IP IP
IP
• FTP
Linux Windows
• HTTP
IP IP IP IP
• NDMP
NFS CIFS
FS
CACHE
POOL
Working Principles of NFS
POR TMAP
Client Server Port mapping
1. RPC request
table
2. RPC response RPC

Client program
Request
RPC
Registration
3. Communication NFS processes:

rpc.nfsd and
Rpc.mountd
NFS share
information
Typical Application of NFS: Shared Storage for Cloud
Computing
⚫ Cloud computing uses the NFS server as the internal shared storage.
Converged storage
NFS client
Cloud computing server

Internal IP
Firewall
External IP network
network Cloud computing server
NFS server
Cloud computing server
Working Principles of CIFS
Client Server
SMB NEGOTIATE Request
Protocol Protocol
SMB NEGOTIATE Response negotiation handshake
SMB SESSION_SETUP Request

Security
SMB SESSION_SETUP Response Session setup authentication
SMB TREECONNECT Request

Tree Share
SMB TREECONNECT Response connection connection
...... Network file

File operation
operation
SMB TREE_DISCONNECT Request
Disconnection Disconnection
SMB TREE_DISCONNECT Response
Typical Application of CIFS: File Sharing Service
⚫ The file sharing service applies to scenarios such as enterprise file servers and
Enterprise office
media assets.
IP Windows IP IP Management
Performance Service Share User

IP IP IP monitoring management management management
NAS service
LAN
DNS
Authentication flow
AD server Management flow
Data flow
Contents
1. DAS
2. NAS
3. SAN
◼ IP SAN Technologies
▫ FC SAN Technologies
▫ Comparison Between IP SAN and FC SAN
NIC + Initiator Software
The initiator software

converts iSCSI packets into
TCP/IP packets, which
consumes host resources.
NIC
TCP/IP-based Ethernet IP SAN

connection
Internal bus
Storage device Ethernet
TOE NIC + Initiator Software
The initiator software
implements the
functions of the iSCSI
layer, which consumes
host resources.
The TOE NIC implements
TCP/IP conversion, which
TOE NIC does not consume host
resources.
TCP/IP-based IP SAN
Ethernet connection
Internal bus
iSCSI HBA
The iSCSI HBA converts

iSCSI packets into TCP/IP
iSCSI HBA packets, which does not
consume host resources.
TCP/IP-based IP SAN
Ethernet connection
Internal bus
Logical Port
⚫ Logical ports are created based on bond ports, VLAN ports, or Ethernet ports. The logical ports
are virtual ports that carry host services.
⚫ A unique IP address is allocated to each logical port for carrying its services.
No. Description
Logical port
Indicates that multiple Ethernet ports are 4
1
bonded to form a bond port.
One to One to
5
Indicates that an Ethernet port is added many many
2
to multiple VLANs.
Indicates that a bond port is added to One to
3
One to
3 Bond port VLAN
multiple VLANs. many many
Indicates that a bond port is used to
4
create multiple logical ports.
Many to One to
2
Indicates that a VLAN port is used to one many 6
5
create multiple logical ports. 1
Indicates that an Ethernet port is used to Ethernet port

6
create multiple logical ports.
VLAN Configuration
⚫ VLAN is a technology that logically divides a physical LAN into multiple broadcast
domains.
⚫ Ethernet ports or bond ports in a storage system can be added to multiple independent
VLANs. You can configure different services in different VLANs to ensure the security
and reliability of service data.
VLAN 1 VLAN 2 VLAN 3 VLAN k

ID = 1 ID = 2 ID = 3 ...... ID = k
Ethernet port
Bond port
P0 P1 P2 P3 Storage device
Adding ports to VLANs
IP Address Failover
⚫ IP address failover indicates that a logical IP address fails over from a faulty port to an available
port. In this way, services are switched from the faulty port to the available port without
interruption. The faulty port can take over services back after being recovered.
⚫ During the IP address failover, services are switched from the faulty port to the available port,
ensuring service continuity and improving reliability of paths for accessing file systems. This
process is transparent to users.
⚫ The essence is a service switchover between ports. The ports can be Ethernet ports, bond ports, or
VLAN ports.
Ethernet Port–based IP Address Failover
⚫ To improve reliability of paths for accessing file systems, you can create logical ports based on
Ethernet ports.
⚫ When the Ethernet port that corresponds to a logical port fails, the system will:
 Locate an available Ethernet

port of the same type.
 Delete the logical port from
the faulty Ethernet port.
Search
 Create the same logical port Logical port A Logical port A
Logical port A
(IP address a) (IP address a)
on the available Ethernet port Port fault Port switchover
to carry services.
 Ensure service continuity.
Ethernet port A Ethernet port A Ethernet port B
Bond Port–based IP Address Failover
⚫ To improve reliability of paths for accessing file systems, you can bond multiple Ethernet ports to
form a bond port.
⚫ When the Ethernet ports that are used to create the bond port fails, the system will:
 Locate an available port.

 Delete the logical port created
High-speed
on the faulty port. data
 Create a logical port with the transmission
Search
same IP address on the Logical port A
Logical port A
Logical port A
available port. Port fault Port switchover
 Switch services to the
available port.
After the faulty port recovers, it Ethernet port A Ethernet port A Ethernet port C
can take over services again. Ethernet port B Ethernet port B Ethernet port D
Bond port A Bond port A Bond port A
VLAN-based IP Address Failover
⚫ You can create VLANs to isolate different services.
⚫ When an Ethernet port on a VLAN fails, the system will:
 Locate an available port of

the same type.
 Delete the logical port from
Search
the faulty port. Logical port A Logical port A
Logical port A
 Create the same logical port Port fault Port switchover
on the available port.

 Switch services to the Ethernet port A Ethernet port A Ethernet port B
available port. VLAN A VLAN A VLAN B
ID = 1 ID = 1 ID = 1
Contents
1. DAS
2. NAS
3. SAN
▫ IP SAN Technologies
◼ FC SAN Technologies
▫ Comparison Between IP SAN and FC SAN
FC HBA
The FC HBA converts SCSI

packets into Fibre Channel
FC HBA packets, which does not
consume host resources.
Fibre Channel
FC SAN
protocol-based
Internal bus
Storage device FC network
FC Network
Zoning
Zone 1: Server 1 can Zone 2: Server 2 can

access Storage 1. access Storage 2.
Server 1 Server 2
Storage 1 Storage 2
Zone 3: Server 3 can

access Storage 1 and
Storage 3 Storage 2.
Server 3
Contents
1. DAS
2. NAS
3. SAN
▫ IP SAN Technologies
▫ FC SAN Technologies
◼ Comparison Between IP SAN and FC SAN
IP SAN and FC SAN
IP SAN FC SAN
iSCSI
FC FC
IP
IP
FC FC
iSCSI
Comparison Between IP SAN and FC SAN
Item IP SAN FC SAN
Dedicated Fibre Channel networks and
Network architecture Existing IP networks
HBAs
Transmission Limited by the maximum transmission
Not limited theoretically
distance distance of optical fibers
Management and Complicated technologies and
As simple as operating IP devices
maintenance management
Compatibility Compatible with all IP network devices Poor compatibility
High purchase (Fibre Channel switches,
Lower purchase and maintenance costs HBAs, Fibre Channel disk arrays, and so
Cost than FC SAN, higher return on on) and maintenance (staff training,
investment (ROI) system configuration and supervision,
and so on) costs
Disaster recovery Local and remote DR available based High hardware and software costs for
(DR) on existing networks at a low cost DR
Security Relatively low Relatively high
Comparison Between DAS, NAS, and SAN
Storage System Architecture DAS NAS SAN
Data transmission protocol SCSI/FC/ATA TCP/IP FC
Transport object Data block File Data block
Using standard file sharing No Yes (NFS/CIFS......) No

protocols
Centralized management Not sure. Yes Management tools required
Improving server efficiency No Yes Yes
Disaster tolerance Low High High, dedicated solution
Application scope SME servers and JBOD SME, monitoring, and Large enterprises and data centers
broadcasting
Application environment LAN LAN Fibre channel storage domain
Documents are seldom shared, the Documents are highly shared, and network
operation platform is independent, different media formats have high Complex network environment,
and the number of servers is storage requirements. high degree of document sharing,
small. heterogeneous operating system
platform, and a large number of
servers.
Capacity expansion capability Low Medium High
Contents
1. DAS
2. NAS
3. SAN
Scale-out Storage Networking
Node 1
Front-end
service network
Node 2
Node 3
Management
network
Node 4
Node 5
Back-end storage network
Networking Overview
⚫ Front-end service/Tenant network
 The front-end service/tenant network is used to interconnect the scale-out storage with the customer network. It provides
the tenant UI for tenant users to complete operations such as resource application and usage query, and processes
service requests sent by tenant clients or APIs.
⚫ Back-end storage network

 The back-end storage/internal management network is used for internal interconnection between nodes. It provides
heartbeat communication between high availability (HA) components such as the data service subsystem (DSS), and
internal communication and data interaction between components.
⚫ Management network
 The management network is used to interconnect with the customer's maintenance network. It provides a management
UI for the system administrator to perform service operations such as system configuration, tenant management,
resource management, and service provisioning, as well as maintenance operations such as alarm, performance, and
topology management. In addition, the Mgmt ports of all physical nodes can be aggregated to provide remote device
maintenance capabilities, such as remotely logging in to the virtual KVM of a device and viewing hardware running data
such as temperature and voltage.
Network Planes
Management Management
Private client Standard protocol node
node FSM
access mode FSM access mode
Storage node Compute node Storage node Compute node

REP REP
VBS Application Application
VBS/HDF
EDS EDS
S-SVR
OSD OSD
Third-place Third-place
DR site quorum DR site quorum Management plane:
server server Storage plane:
Replication plane:
Arbitration plane:
Service plane:
Networking Rules
⚫ Nodes must be placed in a cabinet from bottom to top.
⚫ A deployment solution is usually chosen based on project requirements. The
total power consumption and weight of the storage nodes, switches, and KVM
in a cabinet must be calculated and the number of nodes that can be housed by
a cabinet must be determined based on the equipment room conditions.
⚫ In typical configuration, nodes and switches in the base cabinet are connected
through network cables and SFP+ cables, and nodes in an expansion cabinet
connect to switches in the base cabinet through network cables and optical
fibers.
Quiz
1. Which of the following are included in scale-out storage networking?
A. Management network
B. Front-end service network
C. Front-end storage network
D. Back-end storage network
2. Which of the following protocols are commonly used in SAN networking?

A. FC
B. iSCSI
C. CIFS
D. NFS
Summary
DAS
NAS IP SAN
Storage network
architecture evolution
SAN FC SAN
Comparison between IP SAN and FC

Distributed architecture
SAN
Recommendations

⚫ Popular tools
 HedEx Lite
Acronyms and Abbreviations
Direct Attached Storage (DAS): An external storage device is directly connected to a computer through a cable.
Redundant Array of Independent Disks (RAID): It is a technology that provides a disk group (logical disk) consisting of multiple
disks (physical disks) combined in different modes. The disk group features higher storage performance over a single disk and
supports data redundancy.
Redirect on write (ROW): A core technology used to create file system snapshots. When a source file system receives a write
request to modify existing data, the storage system writes the new data to a new location and directs the BP of the modified
data block to the new location.
Virtual Local Area Network (VLAN): A VLAN is a group of hosts with a common set of requirements that communicate as if they
were attached to the same broadcast domain, regardless of their physical location. VLAN membership can be configured through
software instead of physically relocating devices or connections.
HBA: Host Bus Adapter
KVM: keyboard, video, and mouse. You can use the KVM to remotely view the screen of the terminal host or use the local mouse
and keyboard to remotely control the terminal host. In this way, the administrator can remotely solve the problems that occur on
the terminal host.
intelligent world.


that
Huawei Intelligent Storage Products
Foreword
⚫ This course describes features, positioning, and typical application scenarios

of Huawei intelligent storage products, including Huawei all-flash storage,
hybrid flash storage, scale-out storage, hyper-converged storage, and
backup storage.
Objectives
Upon completion of this course, you will understand the following key
information about Huawei intelligent storage products:
 Features
 Positioning
 Typical Application Scenarios
Contents
1. Panorama
2. All-Flash Storage
3. Hybrid Flash Storage
4. Scale-Out Storage
5. Hyper-Converged Storage
6. Backup Storage
Huawei Storage Products
Solutions
Intelligent data lake Video cloud HPDA Enterprise application acceleration Active-active Backup
Intelligent O&M Data management engine Device management

Data
management
eService DME Storage DeviceManager
All-flash storage Hybrid flash storage Scale-out storage Hyper-converged Backup storage
storage
Storage
products
Contents
1. Panorama
6. Backup Storage
All-Flash Products
High-end
OceanStor Dorado 8000 OceanStor Dorado 18000 V6
Mid-range
OceanStor Dorado 3000 OceanStor Dorado 5000 OceanStor Dorado 6000
Product Features
Ever fast Always-on AI-powered
Industry-high performance SmartMatrix full-mesh Ultimate convergence

Industry-low latency architecture of SAN and NAS
Always-on key services Full-lifecycle intelligent
management
Huawei OceanStor
21 million IOPS Dorado all-flash storage
0.05 ms latency Tolerance of failures of seven
Intelligent O&M
30% higher NAS out of eight controllers
Device-cloud synergy
performance than the Integrated SAN and NAS and
industry average active-active architecture
Device Model Examples
Back panel Front panel
High-end
controller
enclosure
4 U, 28 interconnect I/O modules 4 U, 4 controllers per enclosure
Mid-range
controller
enclosure
2 U, 2 controllers per enclosure, 12 interface modules
2 U, 36 NVMe SSDs (high-density)
Entry-level
controller
enclosure
2 U, 2 controllers per enclosure, 6 interface modules
Smart disk
enclosure 2 U, 25 SAS SSDs
Innovative and Intelligent Hardware Accelerates Critical Paths
(Ever Fast)
Intelligent multi-protocol
1 Transmission interface module
2x reduced network latency
2 Computing High-performance
computing platform
2x improved performance
3 Intelligence Intelligent acceleration

module
50% improved read hit ratio
4 Storage Intelligent SSD

2x reduced write latency
Intelligent
5 Management management hardware
30% improved fault
locating accuracy
Software Architecture
Maintenance terminal software Application server software
SmartKit eService OceanStor BCManager
Management host OS UltraPath SAN host OS layer
Management port/Serial port Storage system software iSCSI/Fibre Channel
Management function
OceanStor DeviceManager F SNMP CLI Syslog
control software
Basic function control software Value-added function control software

Quota
Cache Pool HyperSnap HyperReplication HyperClone HyperCDP HyperMetro
management
File F
sharing F
RAID SCSI SmartThin SmartVirtualization SmartDedupe and SmartCompression
protocol
File system volume
SmartQoS SmartMigration SmartContainer HyperMetro-Inner Three-copy
management module
Storage OS
Intelligent Chips
Controller with0five chips FlashLink® intelligent algorithm

0
00 0 0
1
1
010
11 Reconstruction
0
1
0 10
1
0 Multi-core
1 0
1 00 0 scheduling
0 0 01
1
0
0 101
1 1
X00 minutes/TB 1X minutes/TB Restart count 1 2 3 4 5
1 010 1
1
1 10101enclosure
Smart0disk
0
0
0 Kunpeng chip +
Kunpeng chip + service offloading
AI chip + cache algorithm
1 1010 1 multi-core algorithm Improved read hit ratio
1 0 01
0 0101 0
Faster reconstruction
01
1010 1
1 01 0 01 0
11 0 Data read
0 1 1 10 1 Metadata Data write
0 0
1 00 0 00
1 1 11 New data Advanced feature
0 1 1 11 Disk reconstruction
GC migration
1 10
00 0
1 1 SSD
111 1 data
Garbage collection
1
11 1
0 11 Full-stripe write
Multi-stream data partitioning Global I/O priority adjustment
Reduced write amplification Constant low latency
Reduced garbage collection
1
Typical Application Scenario – Mission-Critical Service
Acceleration
Data center
Operational Product Supervision

Risk control
analytics management and reporting
SAN
Data Data
Data analysis
extraction consolidation
......
Huawei OceanStor Dorado
Contents
1. Panorama
6. Backup Storage
Hybrid Flash Storage
High-end
OceanStor 6810 OceanStor 18510 OceanStor 18810
Mid-range
OceanStor 5310 OceanStor 5510 OceanStor 5610
New-Gen OceanStor Hybrid Flash Storage
Fully upgraded to provide ultimate efficiency

• SmartMatrix active-active architecture with load balancing,
maximizing performance
• Adaptive layout of hot data, improving performance by 100%
Diversified future-oriented functions

• Compatible with different application ecosystems and supporting various
cloudification routes
• Multiple security protection methods, protecting valuable enterprise data
Simpler, smarter, and more economical

• Cross-generation device interconnection and reuse, SAS HDDs not required,
reducing construction costs
• Intelligent and automatic O&M, greatly reducing costs
2 U: Huawei new-gen OceanStor 2 U: Huawei new-gen OceanStor
hybrid flash storage 5310 hybrid flash storage 5510/5610
4 U: Huawei new-gen OceanStor 4 U: Huawei new-gen OceanStor

hybrid flash storage 6810 hybrid flash storage 18510/18810
Converged SAN and NAS
LUN/File system data

iSCSI/FC NFS/CIFS
Block service File service
RAID 2.0+
Storage pool
SSD cache/Tier NL-SAS tier
Convergence of SAN and NAS resources of the Huawei Intelligent tiering cold and
new-gen OceanStor hybrid flash series hot data of SAN and NAS
Support for Multiple Service Scenarios
Cloud
computing Finance
Carrier
Social
intercourse
Energy
Big data Government
Enterprise
Backup
and DR
Application Scenario – Active-Active Data Centers
EDA
PACS CAD
File sharing
New-gen OceanStor New-gen OceanStor

hybrid flash storage hybrid flash storage
HyperMetro
Gateway-free active-active Production center B
Production center A
HyperReplication HyperReplication
OceanStor storage systems
DR center
Contents
1. Panorama
4. Scale-out Storage
6. Backup Storage
Scale-Out Storage
Performance-oriented
OceanStor Pacific 9950
Capacity-oriented

OceanStor Pacific 9520 OceanStor Pacific 9540
Appearance of OceanStor Pacific 9950
Node
Front view (physical) Rear view (physical)
Data cluster
module
Half-palm
NVMe SSD
Service interface
module
BBU System power

module
1 Front panel cover 2 System subrack

3 BBU 4 Node
5 Half-palm NVMe SSD 6 Fan module
7 Power module 8 Interface module

Components
9 Data cluster module
(physical)
Appearance of OceanStor Pacific 9550
Front view (without the cover) Top view
Rear Components (physical)
Fans
Fans
Fans
Fans
Fans
4/9
0/5
1/6
2/7
3/8
Power Power Power Power
supply 0 supply 1 supply 2 supply 3
Rear View Front panel
1 2 Power module
cover
Expansion Half-palm NVMe
3 4
Status indicators module SSD
Interface module 5 BBU 6 Fan module
7 Chassis 8 3.5-inch HDD
Front 9 Node
Node 0 Node 1
Overall Software Architecture
REST/SNMP
POSIX/MPI NFS/CIFS S3 Storage interface
layer
HDFS SCSI/iSCSI
Storage
management
Installation and
Data protection service deployment
Service
Resource tuning service Storage service configuration
layer User
management
File service Object service HDFS service
Block service Device
Distributed Relational Object Service (DROS) layer management
Alarm
Data redundancy
Cluster status control Distributed data routing Monitoring
management Storage engine
layer Upgrade
Data reconstruction
Strong-consistency replication protocol
and balancing Capacity
expansion
Product Features
OceanStor Pacific series

8K
intelligent scale-out storage
HPC AR/VR 5G Telemedicine
video
• High efficiency
Industry application Government | Finance | Carrier | Smart City | Large enterprise
Multi-protocol interworking and
zero data migration
Data enablement Data integration | Data governance | Data development | Data service
• Secure storage
Data processing High-performance computing Big data Artificial intelligence Solution-, system-, device-, and
Full I/O-level HA designs, ensuring
stack Data storage
24/7 service continuity
NAS Object HDFS Block
Diversified computing capabilities x86 ARM NPU GPU

• Cost reduction
Full-stack optimization from
hardware to solutions and from
Data connection algorithms to the architecture,
reducing TCO by more than 30%
FlashLink - Multi-core Technology
CPU grouping algorithm CPU core-based algorithm Intelligent I/O scheduling
Read/ Data exchange Protocol Data Data read/write

Write I/O channel parsing flushing Read I/O 1 Read I/O 2 Write I/O 1 Write I/O 2
Advanced feature
Batch cache write

Core Core Core Core Core Core Core Core
Disk reconstruction
Dedicated Dedicated Shared Garbage collection
I/O grouping to avoid mutual interference Continuous I/O execution to avoid I/O priority guarantee, reducing impact
Dedicated cores are used to ensure the key service consumption caused by switchover on services
resource investment and reduce the latency. A request is continuously processed on the Data read and write I/Os always have the
Shared cores are used to balance the load of same core to prevent thread switchovers for highest priority to ensure the lowest latency.
multiple services. atomic and lock-free operations. This avoids
frequent multi-core switchovers and improves
the CPU cache hit ratio.
Continuous low latency ensures fast response of mission-critical services.

SmartInterworking
Improved analysis efficiency
Data collection Data analysis Data publishing
Multiple protocols share the same data without migration required.
Zero copy Zero copy
Saved storage space
Repeated data storage is avoided, saving space overhead.
NFS/SMB/POSIX/MPI-I/O HDFS S3 Product P of Product E of Product H of OceanStor

Specifications
Vendor D Vendor D Vendor H Pacific
Big data Object

Native semantics √ ╳ ╳ √
File service
service service Gateway-free √ ╳ √ √
Medium Low Low High
Semantic integrity
OceanStor Pacific Impact on

Medium High Medium Low
performance
Cross-protocol
real-time access √ √ ╳ √
Typical scenarios: autonomous driving, oil and gas Multi-protocol
exploration, carrier operation analysis, and log retention sharing ╳ ╳ ╳ √
SmartIndexing
Single-protocol metadata indexing Unified metadata indexing
Single-protocol query Unified query of files,

objects, and big data
Object
Metadata Metadata
Data blocks File metadata indexing engine Data blocks Unified metadata
Periodic Proactive indexing engine
scanning reporting
Message queue
Multi-protocol unified search Results returned in seconds

Proactive reporting of updates
Unified metadata for object, file, and Proactive reporting of metadata updates, Customized search for hundreds of
big data storage improving intelligence billions of metadata records
Parallel File System
Architecture: distributed
lock-free design
⚫ Hybrid load-based design DHT ring .... N1
Metadata is scattered and
N7 N2
owned based on directories
N6 N3
and small I/O forwarding is
Seismic data Seismic data Seismic data Seismic data
processed by the owning
N5 N4
processing interpretation processing interpretation

nodes, eliminating distributed
lock overheads and greatly
... ... reducing the read/write
latency of small I/Os.
Client
I/O: large I/O passthrough
POSIX/MPI NFS/SMB Large I/O Small I/O
and small I/O aggregation
(> 256 KB) (< 256 KB)
L1 Large I/Os are written to the
POSIX/MPI NFS/SMB I/O persistence layer to reduce
passthrough Cache
forwarding. Small I/Os are
Data copy written after aggregation at the
cache layer, reducing the number
of I/O interactions and latency.
Vendor D Vendor N OceanStor Pacific storage 2 MB Disk: multi-granularity disk
Over 10 PB data Over 500 TB data EB-level scalability space management
Over 50 GB/s 600,000+ IOPS 5 U 160 GB/s bandwidth Two-level indexes. The primary
bandwidth requirement requirement and 2M IOPS Index1 Offset1 Length1 index uses the fixed-length
Index2 Offset2 Length2 granularity index to ensure large
One storage system meets the requirements of large files Index3 Offset3 Length3
I/O performance. Sub-indexes are
with high bandwidth, small files with high OPS, and MPI-IO.
Index4 Offset4 Length4
automatically adapted based on
I/O sizes to flexibly cope with
No data copy is required, improving resource utilization. small I/O read and write.
Distributed Parallel Client (DPC)
High single-stream A high-performance client
MPI-IO accelerates
performance unleashes the potentials
meteorological data
ensures the receival of of a fat client.
analysis and application.
satellite data.
NFS client access mode DPC client access mode Maximum single-stream
bandwidth (GB/s)
Application
10
Posix/MPI-IO 8
NFS 8
DPC 6
client DPC 6
4
TCP RDMA 2
2
0
Storage node Storage node Storage node Storage node Storage node Storage node EMC(NFS) DDN(Posix) Huawei(Posix)
⚫ A single client connects to one Up to 8 GB/s single-stream

⚫ A single client connects to multiple bandwidth, which is leading
storage node. storage nodes.
⚫ MPI-IO is not supported. vs. ⚫ MPI-IO is supported.
in the industry
and more than four times
⚫ TCP network access is supported. ⚫ RDMA network access is supported. that of the NFS protocol.
Intelligent Stripe Aggregation
File 1 File N
...
A1 A2 A3 A4 ...
B1 B2 B3 B4 ...
Large I/O passthrough
A1, B1, A3, B6, B9, A5, B5, and A8 of files 1 and N Large I/Os form EC stripes and are directly written
are randomly written. to disks without being cached, saving cache
Intelligent ... resources and prolonging the cache service life.
A1 B1 A3 B6 B9 A5 B5 A8
aggregation
L1 cache
New stripe 1 A1 B1 A3 B6 P Q
Intelligent small I/O aggregation
New stripe 2 B9 A5 B5 A8 P Q ... I/Os are aggregated into full EC stripes and then
written to disks to reduce the write latency.
Random small I/Os are
aggregated into large I/Os.
Full stripe Full stripes are written to disks.
Intelligent append write
A1 B1 A3 B6 P Q Append writes replace traditional overwrite
operations. Random writes change to 100%
B9 A5 B5 A8 P Q
... ... ... ... ... ... sequential writes. Small block write is avoid,
and the write performance of HDDs is
Node1 Node2 Node3 Node4 Node5 Node6 maximized.
Solutions for Four Typical Scenarios
Virtualization and cloud
HPC
resource pool
Government, education, automobile
enterprise, and life science Government, finance, and carrier
HPC parallel storage pool, providing Virtualized storage resource pool
leading performance and elastic scalability and elastic scaling
Decoupled storage and

compute for big data
Carrier, public safety, and
government
Cost-effective big data storage with
decoupled storage and compute
resources
Object resource pool
Finance and government affairs
Hundreds of billions of objects allowed
in a bucket and EB-level scalability
OceanStor Pacific series

Contents
1. Panorama
6. Backup Storage
Hyper-Converged Series
Virtualization/VDI Database Enterprise branch
42 U
24 U
8U
FusionCube 1000H FusionCube 1000D FusionCube 1000C
FusionCube Scenarios and Solutions
Enterprise Intelligent
Virtualization VDI Database
branch campus
Enterprise
Data center solution
branch solution
42 U
24 U
x86 + Kunpeng
Dual ecosystems 8U
1 U/2 U dedicated 5 U dedicated

hardware Cabinet hardware
Logical Architecture of the Hyper-converged Data Center Solution
Edge computing Virtualization VDI Citrix ERP Container Oracle GaussDB
Huawei FusionCompute VMware vSphere RHEL OS

(KVM) (ESXi) physically deployed
virtualization platform virtualization platform database
FusionCube 1000
O&M management
Deduplication and Erasure code Asynchronous
Active-active QoS
FusionCube Vision Pro compression (EC) replication
Multiple pieces of HCI
management software Distributed Multiple
Thin provisioning Linked clone Snapshot
cache resource pools
FusionCube Vision
HCI management
software x86 platform Kunpeng platform GE/10GE/25GE/100GE NVMe SSD GPU/NPU
Ethernet/IB/RoCE
Highlights
Convergence and efficiency Rock-solid reliability Cloud-based and easy-to-use
Convergence of virtualization, container, Ten-level protection for devices, in data Four-level management:
database, desktop cloud, and storage centers, and between data centers Hyper-convergence, multi-site,
The minimum configuration requirement Subhealth detection and end-to-end DIF resource pool, and service-oriented
is lowered to two nodes. ensure ultimate data reliability. Out-of-the-box hyper-convergence
EC, deduplication, and compression improve The real active-active architecture ensures and one-click delivery
the capacity utilization by three times. 99.9999% reliability and service continuity. Unified management platform and
simplified O&M
Four types of services

Ten-level protection ensures Cloud-based and easy-to-use four-
provide convergency
ultimate reliability. level management is supported.
and efficiency.
Enterprise Branch: Easy Storage, Unified Management, and
Cloud Synergy
Easy storage Unified management Cloud synergy
Fast deployment | Plug-and-play Unified management | Algorithm collaboration |
| One-click configuration Centralized monitoring Data collaboration
FusionCube
Vision Pro
FusionCube
Storage | Compute | AI | Network | Security Device | VM | Container | Application Inference | Training
Delivery period shortened by 50% Unified management of one set of software Intelligent autonomy
• Central training and near-end inference

• Application, VM, container, server, firewall,
• Full-stack design and one-stop delivery switch, gateway, storage device, UPS, and sensor • Cloud-based training and near-end inference
• Plug-and-play and one-click configuration
• Visualized remote management of 20,000+ sites • Centralized data backup and fast restoration
Contents
1. Panorama
6. Backup Storage
Backup Storage
Rapid backup and restoration

... End-to-end backup Instant recovery
Backup media Backup
acceleration
servers server
Efficient reduction
IP (recommended)/FC Backup data preprocessing Compression and
and variable-length compaction
deduplication
High reliability
Data consistency System-level Proactive
verification reliability predictive O&M
OceanProtect backup storage
Rear panel Front panel
OceanProtect X9000
storage controller enclosure
4 U, 28 interface modules 4 U, 4 controllers per enclosure
OceanProtect
X6000/X8000 storage
controller enclosure
2 U, 2 controllers per enclosure, 2 U, 25 SAS SSDs
12 interface modules
SAS SSD disk enclosure

2 U, 25 SAS SSDs
NL-SAS disk enclosure
4 U, 24 NL-SAS disks
Hardware Architecture
Power Management
module module
Interface module
Power-BBU module
System subrack Power-BBU module
Controller
BBU
Controller
Controller System subrack

System subrack
Fan module
Disk module Disk module
OceanProtect X9000 OceanProtect X8000 OceanProtect X6000
Software Architecture
Maintenance terminal software Application
SmartKit Backup server
eService Application server
Management port/Serial port Storage system software iSCSI/Fibre Channel
Management function
DeviceManagerF SNMP CLI Syslog
control software
Basic function control software
Cache Pool SCSI HyperSnap HyperReplication
F
RAID File sharing protocol Quota management SmartQoS SmartThin
SmartDedupe and SmartCompression
All-Scenario Data Protection for the Intelligent World
Service Production and Report Data Development Images Medical Regulatory
continuity transaction ERP recoverability compliance
analysis and testing and videos archives
Full DR of hot data Hot backup of warm data Warm archiving of cold data
Data availability Data durability
Data backup
Intelligent DR Service host OceanProtect

management backup storage Data archiving
OceanStor BCManager
Blu-ray archiving
DPA Archiving software + Blu-ray
data protection appliance optical disc library
Active-active/Active-passive/3DC
OceanStor storage
E2E Backup Acceleration Improves System Performance
Front-end network optimization
Backup Full DTOE protocol stack offload

system PHYh
MAC I/O
IP
TCP DTOE NIC
Buffer
➢ Industry-leading 10GE/25GE/100GE Driver File system
OS
OceanProtect backup storage 10/25/100 ➢ Parses the TCP/IP protocol and releases Protocol
Gbit/s DTOE Backup data User space
CPU resources.
Smart NIC Smart NIC
Backup and restoration bandwidth
CPU computing optimization Unit: TB/hour 172
enclosure A 155
enclosure B ➢ Intelligent CPU partitioning and
grouping scheduling algorithm
➢ CPU multi-core load balancing
optimization
➢ Sequential large I/O optimization
Storage pool Disk for backup services
Disk enclosure 1 41
enclosure 1
Back-end network optimization 19 22 18
Flash-based 15
6
Disk Disk ➢ High-performance SSDs and hardware
enclosure 2 enclosure 2
acceleration reduce the I/O latency. Mid-range High-end
➢ ROW full-stripe write reduces write
Backup of peer vendors
amplification.
Backup of Huawei
Throughout-Data-Flow Deduplication and Compression
Implement Ultimate Data Reduction
Periodic backup Data for
Data header
deduplication Data of different
applications
X A B C
D E F
Fingerprint calculation,
Changed data Multi-layer variable- search, and comparison Data is written to
length segmentation Dedicated compression disks after being
algorithm for backup compressed.
Algorithms as the key Backup data Multi-layer inline variable- Feature-based Byte-level
to data compression preprocessing length deduplication compression compaction
Global Algorithm Competence Center Up to 72:1 data compression ratio
Paris Hangzhou
Data deduplication algorithm Data compression algorithm

Before data After data
reduction reduction
The Active-Active Architecture and RAID-TP Ensure System-
level Reliability of Services and Data
Dual-controller active-active Disk-level reliability
RAID-TP: Tolerates simultaneous

failures of three disks without
interrupting services.
Online service Load balancing

switchover
If a controller is faulty, ongoing Symmetric client access and
jobs can be switched over data balancing among all
within seconds without controllers
affecting backup services.
Application Scenario: The Standard NAS/SAN Protocol Is
Compatible with the Backup Software Ecosystem
......
Mainstream backup system
CIFS/NFS/FC SAN/IP SAN
Production system
OceanProtect X8000 OceanProtect X9000 OceanProtect X6000

disk and controller disk and controller disk and controller
integration separation integration
Quiz
1. (Single-answer question) Which of the following statements about Huawei OceanStor all-flash
storage is incorrect?
A. Supports SSDs and NVMe SSDs.
B. Supports SAS disks.

C. Supports HDDs.
D. Supports palm-sized SSDs.
2. (Multiple-answer question) Which of the following storage services are supported by Huawei scale-
out storage?
A. Block storage
B. File storage
C. Object storage
D. HDFS storage
E. Linked storage
Summary
All-flash storage
Hybrid flash
storage
Introduction to Huawei Features, positioning, and

Scale-out storage application scenarios
intelligent storage products
Hyper-converged
storage
Backup storage
Recommendations

 Enterprise service: https://e.huawei.com/en/
⚫ Popular tools
 HedEx Lite
DME: Data Management Engine
HPDA: High Performance Data Analytics
RAID: Redundant Array of Independent Disks. It is a technology that provides a disk group (logical disks) consisting of multiple
disks (physical disks) combined in different modes. The disk group features higher storage performance over a single disk and
supports data redundancy.
ROW: Redirect on write. A core technology used to implement file system snapshots. When the source file system receives a data
write request and data in the source file system needs to be modified, the storage system specifies a new storage location in the
storage pool for the new data and directs the modified data block to the new storage location.
OLTP: Online transaction processing
OLAP: Online analytical processing
NDMP: Network Data Management Protocol. It is an open protocol for network-based backup of file service systems that allows
platform-independent data storage.
intelligent world.


that
Storage System Operation Management
Foreword
⚫ This course describes three methods of managing storage systems:

OceanStor DeviceManager, common line interface (CLI), and UltraPath, as
well as management content and related operations.
Objectives
Upon completion of this course, you will be able to know:

 DeviceManager, CLI, and UltraPath storage system management tools
 Basic management operations of the storage system
Contents
1. Storage Management Overview
2. Storage Management Tools
3. Basic Management Operations
Storage Management Definition
⚫ Storage management allows users to use management tools to query, set, manage,
and maintain storage systems.
Storage resource allocation
Management User management

content
Data protection feature management
Device performance monitoring
Alarm management
......
Common Storage System Access Mode
In what ways can I log in to a

storage system?
Log in to DeviceManager.
Log in to the CLI.
Introduction to DeviceManager
⚫ DeviceManager is storage management software designed by Huawei for a single storage system. It
can help you easily configure, manage, and maintain storage devices.
⚫ Main software functions include storage resource allocation, user management, data protection feature
management, device performance monitoring, and alarm management.
Logging In Using DeviceManager
You must add port number 8088 after the IP

Enter an IP address. Enter port number 8088. address of the management network port.
Otherwise, the login fails.
Format: https://xxx.xxx.xxx.xxx:8088
Enter the user name.

Enter the
password.
Introduction to the CLI
⚫ CLI allows you to manage and maintain the storage system. Configuration commands are entered on the keyboard and compiled
and executed by programs. The command output is displayed in text or graphic format on the CLI.
⚫ Terminal software is required for logging in to the CLI. PuTTY is used as an example.
Enter the IP address of the

management network port.
Set Connection type to SSH.
Click Open. The CLI window is displayed

asking you to log in.
Logging In Using the CLI
⚫ Enter the user name and password as prompted. The system asks you to change the password upon the first login. Change the
password immediately to ensure system security. The following information is displayed when the login is successful:
1. Enter the
user name.
2. Enter the
password.
3. Change the
password.
Introduction to UltraPath
⚫ OceanStor UltraPath is the multipathing software developed by Huawei. Its functions include masking of redundant LUNs, optimum
path selection, I/O load balancing, and failover and failback. These functions enable your storage network to be intelligent, stable,
and fast.
A piece of multipathing software installed

on hosts to improve performance and
Overview availability
Controls access to storage devices. UltraPath Runs in kernel mode of the operating
system as a driver of the operating system.
Selects paths between hosts and storage
devices. Supports querying and setting the
operating parameters of driver software in
Improves the reliability of the paths user mode of the operating system.
between hosts and storage devices.
Functions Has different installation programs or
Environment scripts in different operating systems.
Contents
DeviceManager GUI (1)
Note: The GUI may vary slightly depending on the product version and model. The actual GUI prevails.
DeviceManager GUI (2)
No. Name Description

1 Function pane Displays available functions related to the current operation.
2 Navigation bar Lists all functional modules of the storage system.
The alarm statistics area displays the number of alarms by severity

Alarm and task statistics and helps you learn about the running status of the system.
3
area The task statistics area displays all the tasks executed by users. You
can check whether the tasks are executed successfully.
In the device management area, you can view and modify device
4 Device management area
information, and power off or restart devices.
The logout and language area provide buttons of logout and
5 Logout and language area language. DeviceManager supports two languages: English and
simplified Chinese.
Managing the Access Permission of a Storage System
• Account policy
• Login policy
Security • Access control

policy • Account audit
adjustment
Ensuring
system security
• IP address or address
• Administrator-----role User Access IP segment list
• Role--operation rights management address • Operation rights of the
control
super administrator
Storage System User Management
⚫ To prevent misoperations from compromising the storage system stability and service data security, the storage
system defines user levels and roles to determine user permission and scope of permission.
Operation permission
Level: super administrator
Role: super administrator
All functional modules
of the system
User A
Creating a user and specifying the

user level and role
Operation permission Access permission

Level: administrator Level: read-only user
Role: built-in role B Role: customized role C
Disk Storage
...... LUN Host ......
domain pool
User B User C
Roles and Permissions of a User
• Full control permissions
Data protection • Local data protection
• Creates users of different levels. 5 administrator • Remote data protection
management
Super
administrator 1 • HyperMetro
management
4 SAN resource •
administrator
Storage pool
management
Security 2 •
•
LUN management
Mapping view
administrator 3 Network •
management
Host management
administrator • Port management
• Security rule management
• Certificate management
• Physical port management
• KMC management
• Logical port management
• Antivirus management
• VLAN management
• Data destruction management
• Failover group management
• Regulatory clock management
Downloading a DeviceManager Demo
Visit Huawei technical support

website: Go to Tools.
support.huawei.com
Click Storage Simulator

Demo and select the Search for Storage Simulator.
simulator of the desired
version for installation.
CLI Format Conventions (1)
⚫ You are required to follow the format conventions when you use the CLI commands. Typical command
formats are:
change storage_pool relocation_speed relocation_speed=?

1 2 3
 First field: operation that you want to perform, for example, change (modify) and show (query)
 Second field: object of an operation, for example, storage_pool (storage pool) and host (host)
 Third field (available only in some commands): object attribute, for example, relocation_speed (migration rate)
 Other fields: other parameters required
CLI Format Conventions (2)
⚫ For example, change user user_name=? { level=? | action=? }
 change user keeps unchanged.
 user_name=?, mandatory; For level=? and action=?, one of them can be selected.
 For parameter level=?, level= remains unchanged. The value of ? must be an optional value, for example, level=admin.
⚫ Correct command example: admin:/>change user user_name=newuser level=admin
Format Description
Boldface The keywords of a command are in boldface.
Italics The arguments of a command line, which will be replaced by actual values, are in italics.
[] Items in square brackets ([ ]) are optional.
{ x | y | ... } Optional items are grouped in braces ({ }) and separated by vertical bars (|). One item must be selected.
[ x | y | ... ] Optional items are grouped in square brackets ([ ]) and separated by vertical bars (|). Only one item or
no item can be selected.
{ x | y | ... } * Optional items are grouped in braces ({ }) and separated by vertical bars (|). At least one item must be
selected, and at most all items can be selected.
[ x | y | ... ] * Optional items are grouped in square brackets ([ ]) and separated by vertical bars (|). Several items or
no item can be selected.
CLI Command Completion
⚫ On the CLI, you can press Tab or the space bar to use the command completion function.
⚫ The difference between the two keys is as follows: The space key is used to supplement only the current field,
whereas the Tab key is used to supplement all possible values.
Press Tab once to display the available After the starting segment is determined and completed,
starting segments of a command line. press Tab once to display the available adjacent segments
of the starting segment.
admin:/>//Press "Tab"
^ admin:/>add//Press "Tab"
add change create cache_partition clone consistency_group
delete exit export disk_domain host host_group
help import poweroff lun_copy lun_group mapping_view
poweron reboot remove notification port port_group
scan show swap remote_device security_rule smartqos_policy
snmp storage_pool
When all the fields required by the command are entered and the conditions for running the command are met, the
system prompts that the command can be run after you press Tab. In this case, you can press Enter to run the
command.
admin:/>add port ipv4_route eth_port_id=0 type=net target_ip=192.168.3.0 mask=255.255.255.0 gateway=10.0.0.1//Press Tab
Command is executable now.
Context-Sensitive Help
• Enter a question mark (?) to

• Press Ctrl+A to view the query the basic instruction of CLI
optional values of certain operations and detailed
parameters in certain description of command
commands. Generally, these parameters.
Ctrl+A ?
values need to be obtained • After entering the first field of the
from the system. Context- command and a space, enter a
sensitive question mark (?). You can query all
available next fields and the detailed
Help description of each field.
• You can turn pages on the PageDown
context-sensitive help page. PageUp q
• Exit the context-
sensitive help page.
CLI Command Filtering
Purpose Redundant information is deleted, and valid

content is displayed as required.
Filtering
How to Use After entering the complete query command,
function enter | and press Tab or the space bar.
Related • filterColumn column filtering command

Commands
• filterRow row filtering command
CLI Column Filtering Command - filterColumn
show xxx|filterColumn { exclude | include } columnList=?
Columns available
Choose either.
for filtering
exclude: Filter out information that

does not need to be displayed. If multiple columns are involved,
Include: Only the columns to be they are separated by commas (,).
displayed are reserved.
admin:/>show bbu general|filterColumn exclude//Press "Tab"

<columnList=?> column list separated by comma, select one or more admin:/>show bbu general |filterColumn include
separated by comma, the spaces are replaced with \s in the columnList=Inter\sID,ID
parameter list.
columnList=Inter\sID columnList=ID Inter ID ID
columnList=Health\sStatus columnList=Running\sStatus -------- ------
columnList=Current\sVoltage(V) columnList=Number\sOf\sDischarges 0.0A.0 CTE0.0
columnList=Firmware\sVersion columnList=Delivered\sOn 0.0A.1 CTE0.1
columnList=Owning\sController columnList=Electronic\sLabel
CLI Row Filtering Command - filterRow
show xxx|filterRow column=? predict=? [ predict2=? ] value=? [ logicOp=? ]
Columns required Filtering Filtering Logical relationship

Value
for filtering condition condition 2 between columns
predict=? logicOp=?
• not: The logicOp is not. • and: Multiple columns that
• equal_to: a value equal to value=? meet the condition are
• greater_than: a value greater than displayed.
value=? • or: Any column that meets
• greater_equal: a value equal to or the condition is displayed.
greater than value=?
• less_than: a value less than value=?
• less_equal: a value less than or equal
to value=?
• match: regular expression matching
value=?
Error Prompt Function
Purpose Specify the position of the input error in the command and
provide the correct field for reference.
Error
prompt How to Use When the format of the entered command is incorrect, the
system displays the error location with symbol ^.
function
When multiple errors occur in the command, the system

Note displays only the first error.
admin:/>add part
^
port port_group
admin:/>add part
Major Functions of UltraPath
Virtual LUNs mask physical
LUNs and are visible to
upper-layer users.
Read/Write operations are
performed on virtual LUNs.
Mainstream clustering software:
MSS MSCS, VCS, HACMP, Oracle
The path to the owning
Virtual
RAC, and so on
controller of a LUN is used LUN Mainstream database software:
to achieve the best generation Oracle, DB2, MySQL, Sybase,
performance. Optimum Application Informix, and so on
path software
selection support
UltraPath
Failover Failback
Failover occurs if a I/O After link recovery,
link becomes faulty, load failback immediately
preventing service occurs without manual
interruption.
balancing
intervention or service
interruption.
Multiple paths are
automatically selected to
deliver I/Os, improving I/O
performance. Paths are
selected based on the
path workload.
Positioning of Multipathing Software
Without multipathing software With multipathing software

Basic Eliminating
Server Server
Multipathing software function single points of failure
HBA
+ HBA
Single point of failure
A single point of failure means that the
failure of a component in a network
may cause network breakdown. To
prevent single points of failure, high-
LUN LUN LUN reliability systems implement redundant
LUN LUN LUN backup for devices that may suffer
single points of failure and adopt a
cross cable connection method to
achieve optimal reliability. Moreover,
redundant paths assist in achieving
higher performance.
Storage system Storage system
Without multipathing software With multipathing software Basic Load balancing

function
Server Server
Multipathing software
Load balancing
Link Load balancing is another
critical function of multipathing
I/O Link Balanced loads software. With load balancing,
bottleneck and enhanced the system uses the bandwidth
performance
of multiple links to improve
overall throughput.
Common load balancing
algorithms include round-robin,
LUN LUN LUN LUN LUN LUN minimum queue depth, and
minimum task.
Storage system Storage system
Positioning
Server
Application
UltraPath is a type of filter driver software running in the host kernel. It can block
vdisk
and process disk creation/deletion and I/O delivery of operating systems. UltraPath
HBA HBA
UltraPath implements a reliable use of redundant paths. If a path fails or cannot
meet the performance requirement, UltraPath automatically and transparently
SAN
transfers I/Os to other available paths to ensure that I/Os are transmitted
effectively and reliably. As shown in the figure on the right, UltraPath can handle
Storage
many faults such as HBA faults, link faults, and controller faults. Controller A Controller B
Basic Function Severity Description

If a path is faulty, I/Os on the path are automatically transferred to
Failover High
another available path.
After the faulty path recovers, I/Os are automatically transferred
Failback High
back to the path.
The bandwidths of multiple links are used, improving the overall
Load balancing High
system throughput.
Active-Active Architecture with Full Load Balancing in
OceanStor V6
Disk
Even distribution of unhomed LUNs

UltraPath Data on LUNs is divided into 64 MB slices. The slices are
distributed to different virtual nodes based on the hash
I/O I/O I/O I/O result (LUN ID + LBA).
I/O I/O I/O I/O
Host
Load balancing
UltraPath interacts and negotiates with the storage
system, calculates the hash result (LUN ID + LBA) for
each delivered I/O, obtains the corresponding virtual
Storage node, and selects the physical path corresponding to the
virtual node to deliver the I/O. This reduces cross-CPU
vNode 0 vNode 1 vNode 2 vNode 3 distribution in a storage system and improves end-to-
end performance.
LUN
Contents
Basic Management Operations
Obtaining the version

information and ESN
Managing a storage 3 Managing alarms and

system license events
2 4
1 5
Collecting storage
Managing basic system information
information about a
storage system
Configuring basic
storage services
Managing Basic Information About a Storage System - Setting the
Device Time (1)
A protocol used to synchronize the system time of a computer to the

Coordinated Universal Time (UTC). A server that supports the NTP
Introduction protocol is called an NTP server.
NTP Function When alarms are generated, users can accurately determine the alarm
generation time based on alarm logs.
• Set the time manually.

Setting mode • Synchronize with the client time.
• Set automatic NTP synchronization.
Managing Basic Information About a Storage System - Setting the
Device Time (2)
• Managing the device time on DeviceManager
⚫ Managing the device time on the CLI

 The change ntp_server config command is used to automatically synchronize the storage system time with the
NTP server time.
 The show system ntp command is used to query NTP settings.
 The show ntp status command is used to query the NTP status.
 The show ntp_server general command is used to query the settings of the time synchronization function.
Managing Device Licenses (1)
Permission credentials for using various value-added features (such as
Introduction snapshot, remote replication, clone, and SmartQoS)
License
file During routine device management, you need to check whether the
Precautions
license file is available.
Using DeviceManager to manage licenses
Depending on whether a license has been imported or activated, the license operation displayed
in the License Management area can be Import License, Activate License, or Update License.
For an activated license file, DeviceManager provides two control modes:

• Running time-based control: displays the expiration time of the license.
• Capacity-based control: displays the used/total capacity of the license.
Managing Device Licenses (2)
Using CLI to manage licenses
• The export license command is used to export a license file.

Example: export license ip=? user=? password=? license_path=? [ port=? ] [ protocol=? ]
• The import license command is used to import a license file.

Example: import license ip=? user=? password=? license_path=? [ port=? ] [ protocol=? ]
• The show license command is used to query the function configuration of the imported license
file in the system.
Example: show license
• The show license_active command is used to query information about active licenses.
Example: show license_active
Obtaining the Current Version Information of the Device
Version The matching software version can be accurately determined based
information Function on the system version.
⚫ Obtain the current system version information on ⚫ Log in to the CLI as a super administrator.
DeviceManager.  Run the show system general command.
 Product Version indicates the version of the current
storage system.
Obtaining the Device ESN
Introduction Character string that uniquely identifies a device.
ESN
Application Scenarios such as license application, device repair, and eService
scenarios service configuration
• Obtain the ESN using DeviceManager. • Obtain the ESN using the CLI.
>> Run the show system general command.
Managing Alarms and Events
• Email notification
• Email notification
• SMS notification
• SMS notification
• Syslog notification
• Trap notification
• Trap notification
Configuring the Configuring the
event notification alarm notification
method mode
Managing
alarms and
events
Configuring alarm
masking Configuring alarm
dump
• Alarm masking needs to be enabled or
disabled so that the system does not • When the number of alarms
report or reports alarms to the network exceeds the threshold, the alarms
management system (NMS). will be dumped automatically to
a specific FTP or SFTP server.
Collecting Storage System Information (1)
• Prevent storage system faults and other unpredictable

Purpose disasters from damaging the storage system.
• Know the storage system operating status.
Collecting • Regularly export and securely save the system data for
information How
fault locating and analysis.
System data • Configuration information, system logs, disk logs, and

diagnosis files
Collecting Storage System Information (2)
Collecting storage system configuration data using DeviceManager

On DeviceManager:
• You can export the configuration information to collect the information about the current running status of
the system.
• You can download Recent logs or All logs to collect configuration information, event information, and
debugging logs on the storage device.
• You can download DHA Runtime Log List or HSSD Log List to collect disk run logs, I/O statistics and service
life, and S.M.A.R.T. logs.
• You can export the diagnosis file to collect fault information of the device.
Collecting storage system configuration data using the CLI

Log in to the CLI of the storage system as the super administrator and run the following command to export
the configuration file to an FTP or SFTP server:
export configuration_data ip=? user=? password=? db_file=? [ port=? ] [ protocol=? ] [ clean_device_file=? ]
Configuring Basic Storage Services
• The storage space provided by the storage system is divided into

Configuring multiple LUNs.
basic Function • Map LUNs to an application server.
services
• The application server can use the storage space provided by the
storage system.
Using DeviceManager to configure basic storage services
• Creating a storage pool: DeviceManager allows you to create a storage pool in either recommended or custom mode.
• Allocating storage resources by creating LUN groups or file systems.
Configuring Basic Storage Services Using the CLI
3. Using the storage space on an

application server
⚫ Use the storage space on

2. Setting up connection an application server.
• Create a host.
• Create a host group.
• Configure initiators.
1. Creating storage space
• Create a mapping view.
• Configure connectivity
⚫ Create a disk between a host and a
domain. storage system.
⚫ Create a storage
pool.
⚫ Create a LUN.
⚫ Create a LUN group.
Basic UltraPath Configuration Guide
Using the CLI to configure UltraPath:
For example, in Windows, choose Start > All Programs > UltraPath > upadm or
enter upadm on the CLI of the Windows operating system. Then, run the CLI
commands to configure UltraPath.
Note: Windows is used as an example to explain basic configuration commands for

Huawei UltraPath. The configuration commands for other operating systems are
similar. For details, see the user guide of the corresponding operating system.
UltraPath Parameter Settings in Typical Application Scenarios
In most scenarios, default settings of UltraPath are recommended. In some scenarios, you can configure
UltraPath as instructed by the following:
upadm set workingmode={0|1}
⚫ It specifies the load balancing mode at the storage controller level. 0 indicates load balancing between controllers.
1 indicates load balancing within a controller.
⚫ The default setting is load balancing within a controller. UltraPath selects paths to deliver I/Os based on the owning
controller of each LUN.
⚫ When the inter-controller load balancing mode is used, UltraPath delivers I/Os to all paths. This increases latency
due to I/O forwarding between controllers.
Typical Scenario Recommended Configuration

The transmission paths between hosts and storage
0: load balancing between controllers
systems become a performance bottleneck.
1: default setting, load balancing within a
Other scenarios
controller
UltraPath Parameter Settings in Typical Application Scenarios
upadm set loadbalancemode={round-robin|min-queue-depth|min-task}
⚫ It specifies the load balancing mode at the link level. The value can be round-robin, min-queue-
depth, and min-task.
⚫ The default algorithm is min-queue-depth. UltraPath selects the path that has the least number of
I/Os from all available paths to deliver I/Os.
⚫ When round-robin is used, UltraPath selects all available paths between the application server and
storage system one by one to deliver I/Os.
⚫ When min-task is used, UltraPath selects the path that has the least I/O data volume from all
available paths to deliver I/Os.
Typical Scenario Recommended Configuration

The service I/O models delivered by hosts have small
round-robin: round robin algorithm
differences and I/Os need to be balanced on each path.
The service I/Os delivered by hosts are large data blocks. min-task: minimum task algorithm
min-queue-depth: default setting, minimum queue depth

Other scenarios
algorithm
Quiz
1. (Single-answer question) The management IP address of a storage device is
192.168.5.12. Engineer A needs to enter ( ) in the address box of the browser to
log in to the storage device.
A. 192.168.5.12
B. http://192.168.5.12
C. https://192.168.5.12
D. https://192.168.5.12:8088
2. (True or false) DeviceManager can monitor the performance of controllers, front-
end ports, and back-end ports. ( )
Summary
Storage Management
DeviceManager
Overview
Storage System Introduction to Storage

CLI
Operation Management Management Tools
Basic Management Operations

UltraPath
of the Storage System
Recommendations
⚫ Popular tools
 HedEx Lite
LUN: Logical Unit Number. It is used to identify a logical unit, which is a device addressed by
SCSI.
SAN: Storage Area Network
VLAN: Virtual Local Area Network. It is a group of hosts with a common set of requirements
that communicate as if they were attached to the same broadcast domain, regardless of their
physical location. VLAN membership can be configured through software instead of physically
relocating devices or connections.
NTP: Network Time Protocol. It is an application layer protocol used to synchronize the time
between the distributed time server and the client.
intelligent world.
Copyright©2022 Huawei Technologies Co., Ltd.


portfolio, new technology, etc. There are a number of factors that
Storage Resource Tuning
Technologies and Applications
Foreword
⚫ This course describes storage resource tuning technologies, including

SmartThin, SmartTier, SmartCache, SmartAcceleration, SmartQoS,
SmartDedupe, SmartCompression, SmartVirtualization, and
SmartMigration, as well as their service characteristics, implementation
principles, and application scenarios.
Objectives
⚫ On completion of this course, you will be able to understand the
service characteristics, implementation principles, and application
scenarios of the following features:
 SmartThin
 SmartTier&SmartCache
 SmartAcceleration
 SmartQoS
 SmartDedupe&SmartCompression
 SmartVirtualization
 SmartMigration
Contents
⚫ SmartThin
⚫ SmartTier&SmartCache
⚫ SmartAcceleration
⚫ SmartQoS
⚫ SmartDedupe&SmartCompression
⚫ SmartVirtualization
⚫ SmartMigration
Overview
⚫ The traditional deployment of a storage system has the following problems:
 Adverse impact or even interruption on services when expanding the storage space
 Uneven storage space utilization
 Low storage efficiency
⚫ SmartThin
 SmartThin uses the on-demand storage space allocation policy to improve storage
resource utilization and meet service requirements to a greater extent.
Thin LUN
⚫ Definition: A thin LUN is a logical disk that can be accessed by hosts. It dynamically
allocates storage resources from the storage pool according to the actual capacity
requirements of users.
Data in the file system
✓ Data collection: From the perspective of
a storage system, a thin LUN is a LUN
that can be mapped to a host.
✓ Full availability: data can be read and
written properly.
Host volume
✓ Dynamic allocation: Resources are
The allocated capacity
allocated while data is being written. equals the actual capacity
used by the user.
Capacity allocation
Storage Virtualization
⚫ Capacity-on-write (COW): Storage space is allocated from engines upon data writes
based on load balancing rules.
⚫ Direct-on-time (DOT): Data reads from and writes to a thin LUN are redirected.
Space occupied Physical storage space

Thin LUN by data (storage pool)
RAID 5
Space
allocation
Host
Space allocation COW

upon data writes Redirection to
the actual
storage location
Working Principles of SmartThin
Logical virtual Actual physical
Server space space
data
DB RAID
server
disk
data
DB RAID
server
space disk
data
DB RAID
server
disk
data
DB RAID
server
Visible storage Storage Disk Storage

space resource pool resource array
SmartThin Read Process
1. A thin LUN receives a 2. Queries the mapping table between
read request from a host. the thin LUN and the storage pool.
3. Confirms that the space is allocated by the storage pool and returns the
data read from the corresponding area in the storage pool to the host.
Data
2 1 D 3
1
2 D D D
1 2 3 D
1 3 5
0000 4
2 4 6
3 5 D
Thin LUN 6
Storage pool
Mapping table
1. The thin LUN receives a read request from the host.

2. Queries the mapping table between the thin LUN and the storage pool.
3. Confirms that the space is not allocated by the pool and returns all zeros to the host.
SmartThin Write Process
1. A thin LUN receives a 2. Queries the mapping table between
write request from a host. the thin LUN and the storage pool.
3. Confirms that the space is allocated by the pool and performs the write process on the corresponding
area in the storage pool. If the write request asks for releasing space, the space is released.
Data
Data
2 1 D 3
1
2 D D
3
1 2 3
1 3 5
4
1. A thin LUN receives a write 2 4 6
5
request from the host.
Thin LUN 6
2. Queries the mapping table between Storage pool
the thin LUN and the storage pool. Mapping table
3. If the space is not allocated by the pool, the storage system allocates the space first. And then
performs write process on the corresponding area in the storage pool. If the write request asks for
releasing space, a message is returned to the host.
Application Scenarios
⚫ SmartThin can help core system services that require high service continuity,
such as bank transaction systems, expand system capacity online without
interrupting ongoing services, such as back transaction systems.
⚫ SmartThin can assist with on-demand physical space allocation for services
where the growth of application system data is hard to be accurately evaluated,
such as email services and web disk services, preventing a space waste.
⚫ SmartThin can assist with physical space contention for mixed services that have
diverse storage requirements, such as carriers' services, to achieve optimized
space configuration.
Contents
⚫ SmartThin
⚫ SmartQoS
⚫ SmartMigration
SmartTier
⚫ Tiered storage migrates hot data, cold data, and data with different values in
use to specific storage media to effectively balance performance.
⚫ SmartTier is also called intelligent data tiering. SmartTier provides intelligent
data storage management. It automatically matches data of different activity
levels with storage media of different characteristics by collecting and analyzing
data activity levels.
Performance tier SCM drives
Hot data
Capacity tier SSDs
Cold data
Dividing Storage Tiers
⚫ In the same storage pool, a storage tier is a collection of storage media with the
same performance. SmartTier divides storage media into performance and
capacity tiers. The performance tier that consists of SCM drives delivers higher
performance than the capacity tier that consists of SSDs. Each storage tier
respectively uses the same type of disks and RAID policy.
Storage Tier Disk Type Disk Feature Data Feature
Very short response time and Suitable for storing

Performance tier SCM drive
high cost per gigabyte. frequently accessed data.
Short response time and Suitable for storing less

Capacity tier SSD
moderate cost per gigabyte. frequently accessed data.
Managing Member Disks
Hot standby
RAID 6: 5+2 Performance tier (SCM)
Hot standby
RAID 6: 7+2 Capacity tier (SSD)
Storage pool (RAID 6)
Data Migration
⚫ SCM drives provide better performance than SSDs. With SmartTier available,
new data from hosts is preferentially written to the performance tier for better
performance. Cold data that is not accessed for a long time is migrated to the
capacity tier in the background.
⚫ Data management and migration policies are as follows:
 Metadata: Metadata for new writes and garbage collection (GC) is preferentially
stored at the performance tier (SCM) to ensure high-performance access in large-
capacity scenarios. Space is allocated from the capacity tier only when the space of
the performance tier is insufficient.
 User data: User data is preferentially stored at the performance tier (SCM). When
the size of data in the performance tier reaches a certain level, cold data is migrated
to the capacity tier (SSD) in the background. This ensures that the performance tier
stores more hot data and metadata, allowing faster data accessibility.
SmartCache
Advantages of SCM:
• Lower latency and higher IOPS than SSDs SmartCache SmartCache
partition partition
• Larger capacity and lower price than DRAM
• Suitable for using as cache
SmartCache pool
• SmartCache:
Based on the short read response time of SCM drives, SmartCache uses SCM drives to compose a SmartCache pool and
caches frequently-read data to the SmartCache pool. This shortens the response time for reading hot data, improving system
performance.
• SmartCache pool:
A SmartCache pool consists of SCM drives and is used as a complement of DRAM cache to store hot data.
• SmartCache partition:
The SmartCache partition is a logical concept based on the SmartCache pool. It is used to store LUNs and file system services.
SmartCache Write Process
Servers 1. After receiving a write I/O request to a LUN or file system
from a server, the storage system sends data to the DRAM
cache.
2. After the data is written to the DRAM cache, an I/O
response is returned to the server.
DRAM cache 3. The DRAM cache sends the data to the storage pool
management module.
SmartCache pool SCM
4. Data is stored on SSDs, and an I/O response is returned.
5. The DRAM cache sends data copies to the SmartCache
pool. After the data is filtered by the cold and hot data
Storage pool SSDs identification algorithm, the identified hot data is written
to the SCM media, and the metadata of the mapping
between the data and SCM media is created in the
memory.
6. Data is cached to the SmartCache pool, and an I/O
Hot data Hot data flow
response is returned.
Non-hot data Non-hot data flow
SmartCache Read Hit
Servers
1. A read I/O request from an application server is first

delivered to the DRAM cache before arriving at LUNs or file
DRAM cache systems.
2. If the requested data is not found in the DRAM cache, the
SmartCache pool SCM
read I/O request is further delivered to the SmartCache
pool.
3. If the requested data is found in the SmartCache pool, the
Storage pool SSDs read I/O request is delivered to SCM drives. Data is read
from the SCM drives and returned to the DRAM cache.
4. The DRAM cache returns the data to the application server.

SmartCache Read Miss
Servers
1. A read I/O request from an application server is first delivered
to the DRAM cache before arriving at LUNs or file systems.
2. If the requested data is not found in the DRAM cache, the
DRAM cache forwards the read I/O request to the
DRAM cache SmartCache pool.
3. If the requested data is not found in the SmartCache pool
SmartCache pool SCM
either, the SmartCache module forwards the read I/O request
to the storage pool management module to read the data
from SSDs.
4. Data is read from the SSDs and returned to the SmartCache
Storage pool SSDs
module.
5. The SmartCache module returns the data to the DRAM cache.
6. The DRAM cache returns the data to the application server.

Highlights
⚫ Dynamic Capacity Expansion
 You can add SCM media to a SmartCache pool for dynamic capacity expansion.
 You are advised to configure SCM resources of the same quantity and capacity for controllers in the same
controller enclosure. This ensures balanced acceleration performance for LUN or file system services of
multiple controllers in the same controller enclosure.
⚫ Flexible Policy Configuration
 You can enable or disable SmartCache for specific LUNs or file systems without interrupting services.
 You can add one or more LUNs or file systems to a specified SmartCache partition to shorten the data read
response time.
⚫ Adaptive Switch
 When detecting that the SmartCache policy is inefficient (for example, when the SmartCache hit ratio is
low or the CPU is busy), the system stops allowing I/Os to enter the SmartCache pool to reduce the impact
on services. In this case, I/Os are not sent to the SmartCache pool and the requested data will not be found
in the SmartCache pool. In addition, to ensure data consistency, data in the SmartCache pool is cleared.
 When detecting a scenario suitable for SmartCache, the system automatically restores to the previous state.
Contents
⚫ SmartThin
⚫ SmartQoS
⚫ SmartMigration
SmartAcceleration
⚫ SSDs at the performance tier are exclusively used ⚫ The performance layer is shared by multiple disk
and cannot be shared by multiple disk domains. domains, improving utilization.
⚫ Performance tier configurations are difficult to ⚫ The performance layer is distributed by the system to
adjust, lacking flexibility. each pool based on the recommended quota by
default, and can be flexibly scaled.
⚫ Tiers and caches are independent of each other,
and physical disks must be planned separately for ⚫ The performance layer converges tiers and caches,
both tiers and caches. and is automatically configured for optimal overall
efficiency.
Traditional SmartTier SmartAcceleration
Hybrid-flash storage pool Hybrid-flash storage pool
SmartTier storage pool
... ...
HDD HDD HDD HDD
Exclusive use of physical disks All-flash
storage Capacity layer Capacity layer
SSD SSD SSD SSD ... pool
Performance layer: Performance layer:
SSD acceleration quota SSD acceleration quota
Performance tier Shared
Shared Shared
HDD HDD HDD HDD ... Performance layer

SSD SSD SSD SSD ...
Capacity tier
Storage Layer
⚫ Random distribution of hot and cold data cannot maximize the characteristics of disks
on various media. SmartAcceleration uses a multi-dimensional adaptive hot and cold
data sensing algorithm to implement a unified performance layer, where hot and cold
data are automatically identified based on their characteristics and then automatically
distributed, achieving optimal distribution of hot and cold data.
Performance SSDs
layer
Capacity NL-SAS disks

layer
Hot data Cold data
Unified Performance Layer That Flexibly Integrates Caches and Tiers
Unified and converged
performance layer
ROW mechanism accelerating
write performance Aggregation Popularity Association
... Cache
Read Write Traffic
Tier Admission Eviction
Read-after-write distribution
acceleration Migration Integration
Read
acceleration Hit rate and performance
2) Small I/O overwrite
acceleration effect
3) Hot data write/read Service scenario 1
5) Read-only hot data
Flexible Scenario-based
6) Associated prefetch Service scenario 2
adaptive cache
Performance ratio adjustment
Cache Tier layer
Hot Cold Cache capacity
4) Cold data 1) Cold data directly
migration over time written to HDDs
LSM Tree/AR
Key range Key range Key range Key range Key range Key range Key range
...
partition 1 partition 2 partition 3 partition 4 partition 5 partition 6 partition n
L0
Tree/Learning Tree
L1
SSD L2
HDD Capacity layer L3
Flow
HDD
B+ Tree
ROW-based Large-Block Sequential Write
Object 0 Object 1 Object 2
... ... ...
4 KB 3 KB 7 KB 4 KB 3 KB 7 KB 4 KB 4 KB 16 KB Logic space
ROW and I/O

aggregation
Log structure A B C D E ...
Data update
Full stripe
A B C D E P Q
CKG
Traditional Mode OceanStor Mode

Configuration Read Count Write Count Total I/Os Configuration Read Count Write Count Total I/Os
RAID-5 2 1 4 (3) RAID-5 0 0 1
RAID-6 3 2 6 (5) RAID-6 0 0 1
RAID-TP 4 3 8 (7) RAID-TP 0 0 1
Working Principles of SmartAcceleration
Write Read
1. Find hot read

DRAM DRAM data in the DRAM. Intelligent access frequency
Write I/O
identification algorithm
distribution
Intelligent access Small I/Os written to the SSD tier Admission: Check whether
frequency identification Modification to hot read data Miss the data is hot enough to
algorithm written to the SSD tier be distributed to SSDs.
SSD
Admission Miss
performance Cache Tier
SSD layer
performance Cache Tier Large I/Os 2. Find hot read 3. Find hot read
layer aggregated data in the SSD data in the SSD tier. Read
Eviction and flushed cache. Miss (including
Regular migration to disks Hot data I/O
prefetch)
HDD capacity HDD capacity

layer layer
Contents
⚫ SmartThin
⚫ SmartQoS
⚫ SmartMigration
Overview
⚫ SmartQoS:
SmartQoS is an intelligent service quality control feature. It dynamically
allocates storage system resources to meet the performance
requirement of certain applications.
⚫ The controlled objects of SmartQoS include LUNs, LUN groups,
hosts, and file systems.
⚫ By function, SmartQoS can be classified into traffic control management,
burst traffic control management, and lower limit guarantee.
⚫ Other characteristics of SmartQoS include multi-dimensional traffic
control, hierarchical policies, and objective distribution.
Upper Limit Traffic Control Management
Application Application
server 1 server 2
1 I/O request 1 I/O request
Number of allocated tokens

(I/O queue processing Storage system
mechanism of applications in
the system)
2
Process the dequeuing
Application I/O queue 1 I/O requests.
... 3
Application I/O queue n

Disk
Burst Traffic Control Management
⚫ When configuring the burst value, you must configure the upper limit. That means the instantaneous
performance value can reach the burst value, but the average performance value cannot exceed the upper limit.
⚫ The service load in phase 1 is lower than the upper limit, and the burst capabilities are accumulated.
⚫ In phase 2, when the service load increases, the performance can exceed the upper limit and reach the burst
threshold. The total burst duration is the smaller value between the accumulated burst duration in phase 1 and
the burst duration configured in the QoS policy.
⚫ In phase 3, once the burst ends, the service performance falls below the upper limit.
Supplement tokens
2
Burst limit
Lend
tokens
3
Upper limit Token Token
bucket C bucket E
1 Store
tokens
Consume tokens
Lower Limit Guarantee
⚫ When multiple services preempt resources and a small number of critical services have no
advantages in resource preemption, the lower limit guarantee function can automatically
suppress non-critical services to reserve resources for critical services.
Users set a lower limit objective for The system sets a minimum
System rating is
LUNs of a few critical services. But performance value for other LUNs—
the LUNs fail to reach the objective. a shared lower limit objective. triggered.
LUNs with performance LUNs with performance LUNs with performance LUNs with performance
failing to reach the lower lower than the minimum slightly exceeding the far beyond the minimum
limit objective performance minimum performance performance
No suppression No suppression Burst prevention Traffic suppression
Disk
Multi-dimensional Traffic Control
⚫ SmartQoS policies provide multi-dimensional processes to simplify user operations:
 Traffic control for LUNs
 Traffic control for LUN groups
 Traffic control for hosts
 Traffic control for file systems
⚫ Notes:
 Only one type of objects can be added to each policy.
 Only one LUN group can be added to each policy.
 A LUN, LUN group, host, or file system can be added to only one policy.
 A LUN can be added to different LUN groups, and different LUN groups can be added to different
policies. In this case, the final traffic control value for the LUN is set to the smallest value to ensure that
all policies are implemented.
 Host policies do not support the lower limit guarantee function.
Hierarchical Management
⚫ Hierarchical policy: System capabilities are classified by user, vStore, or subsidiary to
prevent interference between different dimensions. Multiple common policies can be
added to a hierarchical policy. The total performance of all sub-policies meets the
requirements of the hierarchical policy.
⚫ Common policy: Traffic control policies are configured for different services of a single
user, vStore, or subsidiary.
⚫ Notes:
 A hierarchical policy has all the features of a common policy, and a common policy can be
added as a sub-policy.
 Sub-policies with LUNs and LUN groups can be added to the same hierarchical policy.
 The policy configuration for a LUN/LUN group, a host, and a file system are mutually
exclusive. Therefore, they cannot be added to the same hierarchical or common policy.
 A lower limit guarantee policy cannot be added to a hierarchical policy.
Objective Distribution
⚫ All objects in a SmartQoS policy share the upper limit objectives. The SmartQoS module
periodically collects the performance and requirement statistics of all controlled objects
in a traffic control policy, and distributes the traffic control objective to controlled
objects.
⚫ Currently, the max-min weight allocation algorithm is used. The figure below shows the
overall objective distribution process.
Calculate the
intermediate
Identify valid Calculate the
result of the
distribution final
Collect max. and
Start objects and objective End
information. min. values
accumulate distribution
for the
weights. result.
objective
distribution.
SmartQoS Application Scenarios
⚫ Upper limit traffic control management:
 Suitable to scenarios with mixed services to prevent services from affecting each
other. It can limit the resource usage of low-priority services that have great impacts
on the system.
 Suitable for scenarios where performance restriction is required. For example, when
an online service and a backup service coexist, SmartQoS restricts the maximum
traffic of the backup service to ensure performance of the online service while
ensuring the backup window.
⚫ Burst traffic control management:
 Suitable for latency-sensitive services, such as database services, VM deployment,
and VM startup.
⚫ Lower limit guarantee:
 Ensures service quality (IOPS, BPS, and latency) for services that have a small
quantity but are the most critical.
Configuration Process
Start
Check the SmartQoS license.

Monitor service performance.
Configure SmartQoS.
Create a SmartQoS policy.

End
Main steps
Sub-steps
Contents
⚫ SmartThin
⚫ SmartQoS
⚫ SmartMigration
SmartDedupe
⚫ SmartDedupe is a data deduplication technology developed by Huawei for saving storage space.
It deletes duplicate data in a storage system by searching for duplicate data blocks (generally
greater than 4 KB) and saves only one copy of these blocks in the storage system. This technology
is widely used in network disks, emails, and backup media devices.
⚫ Type of deduplication
 Inline deduplication: Data is deduplicated when being written to storage media.
 Post-process deduplication: Data is first written to persistent storage media and then read for
deduplication.
 Fixed-length deduplication: Data is divided into blocks of a fixed size for deduplication.
 Variable-length deduplication: Data is divided into blocks of variable sizes based on its content for
deduplication. This is mainly used in data backup.
 Similarity-based deduplication: The system divides data into blocks of a fixed size and analyzes the
similarity among the blocks. The system deduplicates the identical data blocks and performs combined or
delta compression on the similar data blocks.
SmartCompression
⚫ SmartCompression is a data compression technology developed by Huawei. It is a
process of encoding information using fewer bits than the original representation.
⚫ Type of compression
 Inline compression: Data is compressed when being written to storage media.
 Post-process compression: Data is first written to storage media and then read for
compression.
 Software compression: Uses the CPU of a system to perform a compression algorithm.
 Hardware compression: The compression algorithm logic is fixed to the hardware device, for
example, the acceleration engine, FPGA, or ASIC of the Arm CPU. The compression API
provided by the hardware device is invoked when the compression algorithm logic is used.
 Lossy compression: A compression mode in which original data cannot be completely restored
after decompression. This mode is commonly used to process audio, video, and images.
 Lossless compression: A compression mode in which original data can be completely restored
after decompression.
Working Principles of Inline Deduplication
Data to be deduplicated
5. For existing blocks, increases the reference
count (RC) and returns the addresses of the
existing blocks. For new blocks, writes them
1. Divides data into blocks. Block 0 Block 1 Block 2 to the storage space.
6. Adds the fingerprints and storage locations

2. Calculates the fingerprints of the new blocks to the fingerprint table.
FP 0 FP 1 FP 2
of the data blocks.
Fingerprint (FP) table
3. Checks whether the FP Data Address RC

fingerprints exist in the FP 0 = FP x Data address 0 +1
fingerprint table. Old Old New
block block block FP 1 = FP y Data address 1 +1
4. If a fingerprint exists in the
fingerprint table, this block FP 2 Data address 2 1
already exists in the system.
If a fingerprint is not found FP x Data address x 1->2
Increases Increases Writes
in the fingerprint table, the the RC. the RC. data. FP y Data address y 1->2
block is new.
Working Principles of Post-processing Similarity-based Deduplication
Block 1 Block 2 Block 3
SFP 0 SFP 1 SFP 2

FP 0 FP 1 FP 2
Similarity-based deduplication opportunity

Storage
table
<LBA> <FP> <SFP>

Same block LBA 0 FP 0 SFP 0
Similar block
New block (same FP)
LBA 1 FP 1 SFP 1
(same SFP)
(byte-by-byte comparison)
LBA 2 FP 2 SFP 2
Increases the Performs delta Fingerprint (FP) table

Writes data. RC. compression.
<FP> Data Address RC
Saves data after delta compression FP 0 Data address 0 2,1
Working Principles of Similarity-based Deduplication in the Background
1. Divides data into blocks. 7. Adds the fingerprint, storage location, and
reference count of the reference block to the
fingerprint table.
2. Calculates the fingerprints and similar
fingerprints (SFPs) of the data blocks and adds
them to the similarity-based deduplication Block 0 Block 1 Block 2
opportunity table.
8. Updates the mapping of each data block.
3. Analyzes the similarity-based deduplication The LBA points to the fingerprint and storage
opportunity table in the background and SFP 0 SFP 1 SFP 2 location of the reference block.
performs similarity-based deduplication on FP 0 FP 1 FP 2
data blocks with the same SFP.
Similarity-based deduplication
4. Selects a reference block and compares other Performs data persistence. opportunity table
blocks with the reference block. SFP 0 -> FP 0, LBA 0, ...
5. If the fingerprint of another block is the same <LBA> <FP>
as that of the reference block and the two
blocks are consistent byte by byte, the block is LBA 0 FP 0,0
duplicate. Increases the reference count of the Reference Same block Similar block
block FP 1 = FP 0 LBA 1 FP 0,0
reference block. SFP 2 = SFP 0
LBA 2 FP 0,1
6. If another block is different from the
<FP> Data Address RC
reference block, but they have the same SFP, Compresses Increases Performs
performs delta compression on this block. data. the RC. delta FP 0 Data address 0 2,1
After delta compression, writes the data to compression.
persistent storage media.
Performs data persistence after delta
compression.
Working Principles of SmartCompression
(LZ77 example)
abcdbbccaaabaeaaabaee
⚫ Assume that a string abcdbbccaaabaeaaabaee is to be compressed, the first 10 characters have
been compressed, and the window is 10 characters.
Start Position of Length of the Last Character in
Window Uncompressed Identical String Compression Code
the Identical String Identical String the Identical String
abcdbbccaa abaeaaabaee ab 0 2 a (0,2,a)
dbbccaaaba eaaabaee null 0 0 e (0,0,e)
bbccaaabae aaabaee aaabae 4 6 e (4,6,e)
⚫ The compression result is (0,2,a)(0,0,e)(4,6,e).
SmartDedupe and SmartCompression Process
App
1. Host data is written 2. A write success is
Application to the cache of the returned to the host.
storage system.
Device
3. Data is divided into blocks of a fixed
length while being written to disks.
4. Fingerprints are
calculated using weak hash.
5. The system FP Data Address RC
compares the
FP fingerprints with FP 0 Data address 0 3
6. If a fingerprint is the same as
those in the that in the fingerprint table, the
FP 1 Data address 1 2
fingerprint table. system reads the existing data
7. If the fingerprints are different or corresponding to the fingerprint
the data is found different in byte- ... ... ...
and compares it with the new
by-byte comparison, the system data byte by byte. If the data is
compresses the new data. the same, the system increases
the reference count.
Compression 8. Compressed data is written to
SSDs. The mapping between the
fingerprint and storage location is
saved in the fingerprint table, and the
reference count is set to 1.
Scenarios Where SmartDedupe and SmartCompression Are
Used Together
⚫ Using both SmartDedupe and SmartCompression can achieve the best space
saving effect.
⚫ The scenarios include:
 VDI and VSI scenarios
 Data testing or development systems
 Storage systems with the file service enabled
 Engineering data systems
Start
Check the license file.
Enable SmartDedupe and

SmartCompression.
End
Contents
⚫ SmartThin
⚫ SmartQoS
⚫ SmartMigration
Concepts Related to SmartVirtualization
⚫ Question: SmartVirtualization allows external LUNs to provide physical storage
space for OceanStor Dorado series storage systems. How can it be
implemented?
⚫ What do "local storage system", "heterogeneous storage system", "external
LUN", "eDevLUN", "online takeover", "offline takeover", and "hosting" mean?
Relationship Between an eDevLUN and an External LUN
⚫ An eDevLUN consists of data and metadata. A mapping is established between data
and metadata.
Data Read Process
⚫ After an external LUN in a
heterogeneous storage system is
hosted using SmartVirtualization,
when an application server sends a
request to read data from the
external LUN, the eDevLUN in the
local storage system receives the
request and reads data from the
external LUN
Data Write Process
⚫ After an external LUN in a
heterogeneous storage system is
hosted using SmartVirtualization,
the data write process is as follows:
Centralized Management of Storage Resources
⚫ If multiple heterogeneous storage systems have been deployed onsite, the following two
challenges may occur:
 Due to incompatibility issues, the multipathing software on an application server may not be compatible
with all heterogeneous storage systems.
 In a certain network environment (such as a Fibre Channel direct-connection network), one application
server can only be connected to one storage system. However, in actual applications, one application
server needs to distribute services to multiple storage systems.
Offline Takeover and Online Takeover
eDevLUNs without masquerading properties eDevLUNs with masquerading properties
Read and write I/Os
Step 2: Disconnect the Step 3: Connect the Step 2: Disconnect the Step 3: Connect the
application server from application server to the application server from the application server to the local
the heterogeneous local storage system and heterogeneous storage system. storage system and then the
storage system to stop enable the local storage Data transfer interruption does application server can identify
data transfer and system to take over not affect service. the masqueraded eDevLUNs.
services. services.
Mapping Mapping
SmartVirtualization SmartVirtualization
Heterogeneous storage system Local storage system Heterogeneous storage system Local storage system
Step 1: Connect the local storage system Step 1: Connect the local storage system to the heterogeneous
to the heterogeneous storage system. storage system to enable the local storage system to take over
services from the heterogeneous storage system.
Service Data Migration from a Legacy Storage System to a New
Storage System
⚫ As services grow continuously, more storage is required for storing increasing data. The legacy storage system cannot provide
satisfactory data storage capacity and performance. In this case, you can purchase a storage system that provides a larger
capacity and better performance to replace the legacy storage system. As software and hardware of the legacy and new
storage systems are different, the services may be interrupted and data may be lost during data migration.
SmartVirtualization can mask the differences between storage systems to map an external LUN of the legacy storage system
to the new storage system (presented as an eDevLUN on the new storage system). Then SmartMigration can be used to
reliably migrate all service data from the legacy storage system to the new storage system while keeping services running.
New storage system Legacy storage system
Cold Data Migration from a New Storage System to a Legacy
Storage System
⚫ After a legacy storage system is replaced with a new storage system, some data in the new storage system is
rarely accessed, which is called cold data. If massive cold data is stored in the new storage system, the
storage resource utilization of the storage system lowers down, causing a waste of storage space. To reduce
operation expenditure (OPEX), SmartVirtualization can work with SmartMigration to migrate the cold data to
the legacy heterogeneous storage system.
New storage system Legacy storage system
Takeover Mode Selection
⚫ SmartVirtualization provides online and offline modes to take over heterogeneous
storage systems. There are three masquerading types for online takeover, which are
basic masquerading, extended masquerading, and third-party masquerading. The
takeover mode depends on the vendors and versions of the heterogeneous storage
systems and multipathing software.
Takeover Mode Masquerading Description
The offline takeover mode is applicable to all compatible

Huawei and third-party heterogeneous storage systems. In
Offline takeover No masquerading
this mode, services running on the related application
servers are stopped temporarily.
The selection of basic masquerading or extended
Basic masquerading
masquerading depends on the vendor and version of the
Online takeover or extended
multipathing software and the versions of Huawei
masquerading
heterogeneous storage systems.
Start
Complete the installation and

deployment.
Configure and use eDevLUNs.
End
Contents
⚫ SmartThin
⚫ SmartQoS
⚫ SmartMigration
Overview
⚫ SmartMigration is a key technology for service migration. Services on a source
LUN can be completely migrated to a target LUN without interrupting host
services. The target LUN can totally replace the source LUN to carry services
after the replication is complete.
What does
"completely"
mean?
Working Principles of SmartMigration
⚫ SmartMigration is leveraged to adjust service performance or upgrade storage
systems by migrating services between LUNs.
⚫ SmartMigration is implemented in two phases:
Service data synchronization LUN information exchange

Start Finish
SmartMigration Service Data Synchronization
⚫ After creating a SmartMigration task, create the pair relationship between a
source LUN and a target LUN.
⚫ Service data synchronization between the source and target LUNs involves
initial synchronization and change synchronization.
A B C
Host
D E F 1 5
G H I
A B C A B C DCL LOG
2
4
D E F Copies full data. D E F Storage
G H I G H I 4 LM
2 2
Source LUN Target LUN 3 3
Source Target
Initial synchronization LUN LUN
Data change synchronization
SmartMigration LUN Information Exchange
⚫ LUN information exchange is used for mappings between LUNs and data
volumes, namely, the exchange between both data volume IDs when the IDs of
source LUNs and target LUNs remain unchanged.
Source Target
The host reads LUN LUN
the source LUN ID.
Before LUN LUN ID Source data volume ID LUN ID Target data volume ID
1
information exchange 0 0 1 1
Mapping
During LUN LUN ID Target data volume ID LUN ID Source data volume ID
2
Exchanges data volume IDs.
After LUN LUN ID Target data volume ID LUN ID Source data volume ID
3
Mapping
SmartMigration Pair Splitting
⚫ Splitting is performed on a single pair. The splitting process includes stopping
service data synchronization between the source LUN and target LUN in a pair
to exchange LUN information, and removing the data migration relationship
after the exchange.
After service data synchronization is stopped
1. Exchanges
How to split
information
multiple pairs?
Pair
Splitting Source Target
LUN LUN
2. Splits the pair.
SmartMigration Status Transition
⚫ Migrating: Data is being synchronized from
the source LUN to the target LUN. Initial
creation
⚫ Normal: Data is successfully synchronized
between the source LUN and the target LUN.
Migrating
⚫ Paused: The pair is in the waiting queue.
⚫ Interrupted: The replication relationship Sync interruption
Paused
Synchronized
between the source LUN and target LUN is
Fault
interrupted due to an I/O error in rectification
Normal
SmartMigration.
Interrupted
⚫ Migrated: Data is successfully synchronized Splitting
between the source LUN and the target LUN,
and the splitting is complete. Migrated
Start
Create a SmartMigration task.
(Consistently) Split a
SmartMigration pair.
End
Summary
SmartThin
SmartTier&SmartCache
SmartAcceleration
Storage Resource Tuning Definition, Principle,

SmartQoS
Technologies and Applications and Application
SmartDedupe&SmartCompression
SmartVirtualization
SmartMigration
Quiz
⚫ (True or false) SmartTier cannot be enabled for a storage pool whose

member disks are of the same type. ( )
⚫ (Multiple-choice) Which of the following migration policies can be set for
LUNs? ( )
 Automatic migration
 Migration to the higher-performance tier
 Migration to the lower-performance tier
 No migration
Quiz
3. (Single-answer question) Which status must a pair be before consistency

splitting during LUN migration? ( )
 Migrating
 Paused
 Normal
 Migrated
Recommendations

 Enterprise business: https://enterprise.huawei.com/en/
 Technical support: https://support.huawei.com/enterprise/
⚫ Popular tools
 HedEx Lite
intelligent world.


that
Storage Data Protection
Technologies and Applications
Foreword
⚫ Traditional data protection solutions focus on periodic data backup.

Therefore, problems such as no backup window, inconsistent data, and
impact on the production system always occur.
⚫ This course describes storage data protection technologies such as
HyperSnap, HyperClone, HyperCDP, and HyperLock.
Objectives
On completion of this course, you will be able to understand the principles,

configuration methods, and application scenarios of the following features:
 HyperSnap
 HyperClone
 HyperCDP
 HyperLock
Contents
1. HyperSnap
2. HyperClone
3. HyperCDP
4. HyperLock
Overview
⚫ HyperSnap:
It is a snapshot feature. A snapshot generated by HyperSnap is a point-in-
time, consistent, and fully usable copy of source data. It is a static image of
the source data at the copy point in time.
⚫ A snapshot can be implemented using the copy-on-write (COW) or

redirect-on-write (ROW) technology.
 COW enables data to be copied in the initial data write process. Data
copy affects write performance of hosts.
 ROW does not copy data. However, after data is overwritten frequently,
data distribution on the source LUN will be damaged, adversely affecting
sequential read performance of hosts.
ROW Principle
Source file system's Snapshot's

mapping table mapping table
L3->P3
(2)
L1->P1 L4->P4 NULL L2->P2
(3)
L0->P0 L2->P2 L2->P5
(1) Redirect-on-write (ROW)
A B C D E F
P0 P1 P2 P3 P4 P5 P6 P7
Working Principles of HyperSnap
⚫ Definition
 A snapshot is a consistent copy of the source data at a certain point in time. After the snapshot is generated, it can be
read by hosts and used as a data backup at a certain point in time.
⚫ Main features
 Instant generation: A storage system can generate a snapshot within a few seconds to obtain the consistent copy of
source data.
 Small storage space occupation: A snapshot is not a full physical data copy, which does not occupy large storage space.
Therefore, a snapshot for a large amount of source data occupies only a small space.
a b c a b c
d e f d e f
8:00 a.m.
g h i g h i
j k l j k l
a b c a b c
d m f d e f
9:00 p.m.
g h n g h i
j k l j k l
LUN Snapshot Principles: Zero Performance Loss
Data requested to be written to L2 of Data requested to be Data requested to be

the source LUN is written to P5. written to L0 of snapshot written to L2 of snapshot
Data requested to be written to L2 of 1 is written to P6. 2 is written to P8.
the source LUN is again written to P7. ⚫ Data requested to be written to L2
of the source LUN is written to a
new space P5. The original space P2
LUN mapping table Snapshot mapping table Snapshot mapping table
is referenced by the snapshot.
⚫ Data requested to be written to L0
of snapshot 1 is written to the new
space P6, bringing no additional
read and write overhead.
⚫ When data is written to L2 of the
source LUN again, the requested
data is written to a new space P7.
L3->P3 L4->P4 L2->P5 L2->P7 L2->P8 The original space P5 is released
L0->P6 L0->P0 L1->P1 L2->P2
because it is not referenced by a
snapshot.
A B C D E F G H I ⚫ A new snapshot 2 is created and
activated.
P0 P1 P2 P3 P4 P5 P6 P7 P8
SSD storage space
File System Snapshot Principles: Zero Performance Loss
Write L2->P5
File system Creating snapshot 1 Write L0->P6
Snapshots do not affect read
File system mapping table Snapshot mapping table File system mapping table
and write performance of
source file systems.
Snapshots are read-only and

have the same read performance
as the source file systems.
L0->P6 L0->P0 L1->P1 L2->P2 L3->P3 L4->P4 L2->P5 L0->P6
Disk space A B C D E F G
P0 P1 P2 P3 P4 P5 P6 P7 P8
Data requested to be written to L0 and L2 of the source file system is written to new
spaces P5 and P6. The original spaces P0 and P2 are referenced by the snapshot.
HyperSnap Principles: Rollback
Data at 10:00 a.m. Data at 11:00 a.m. Virus infection Data at 11:00 a.m.
Restore
10 11 12 Time
Snapshot at 10:00 a.m. Snapshot at 11:00 a.m.
HyperSnap Principles: Snapshot Cascading and Cross-
Level Rollback
⚫ Snapshot cascading: It is a child
Source volume
snapshot of a parent snapshot. The
difference between snapshot duplicates
and snapshot cascading is that the
latter includes the data of its parent
8:00 9:00
snapshot. Other functions are the same
snapshot0 snapshot1
as common snapshots.
10:00 11:00
Snapshot1.snapshot0 Snapshot1.snapshot1 ⚫ Cross-level rollback: Snapshots
sharing the same source volume can
roll back each other regardless of their
cascading levels.
Key Technologies of HyperSnap: Duplicate
How can I obtain multiple

duplicates of the same
Source volume Snapshot snapshot?
8:00
Snapshots are virtual, so

8:00 8:00 8:00
they can be duplicated fast.
Key Technologies of HyperSnap: Rollback Before Write
How can I instantly recover data?
Source volume Snapshot

Snapshot
rollback During the rollback, when a
8:00
host writes data to a source
LUN, the snapshot copies the
data blocks to the source LUN,
and then the host continues to
write data.
When no host reads or writes
data, the snapshot data is
rolled back to the source
volume in sequence.
Application Scenario
Source Snapshot
LUN LUN
Source
LUN
Snapshot LUN Duplicate Duplicate Duplicate Duplicate
1:00
2:00
4:00 3:00
Report Data Data Decision-

generation test analysis making
support
⚫ Continuous data protection ⚫ Data backup and restoration
Required
Start
Optional
1. Checking the availability Check the HyperSnap

of the snapshot function license file.
Performing this operation when a

Create a source LUN storage system does not have the source
(or file system). LUN or file system of a snapshot.
2. Creating a snapshot Create a snapshot.
End
Contents
1. HyperSnap
2. HyperClone
3. HyperCDP
4. HyperLock
Overview
⚫ Definition
 HyperClone creates a full data copy (a target LUN) of a source LUN at a specified point in
time (synchronization start time).
 HyperClone is to create a clone for a file system or a snapshot of the file system at a specific
point in time. After a clone file system has been created, its data (including the dtree
configuration and dtree data) is consistent with that of the parent file system at the
corresponding point in time.
⚫ Features
 A target LUN can be read and written during synchronization.
 Full synchronization and incremental synchronization are supported.
 Forward synchronization and reverse synchronization are supported.
 Consistency groups are supported.
Working Principles of HyperClone
⚫ Definition: Clone is a consistent data copy of a source data at a specific point in time. It functions as a
complete data copy after data synchronization. It serves as a data backup and is accessible to hosts.
⚫ Main features
 Quick clone generation: A storage system can generate a clone within several seconds to obtain a consistency copy of a
source data. The generated clone can be read and written immediately. Users can configure different deduplication and
compression attributes for the generated clone.
 Online splitting: A split can be performed to cancel the association between a source LUN and a clone LUN without
interrupting services. The split read and write operation on the clone LUN will not affect the I/O process of the source LUN.
Create Synchronize
HyperClone. HyperClone.
a b c a b c a b c
d e f d e f d e f
g h i g h i g h i
j k l j k l j k l
HyperClone Principles: Synchronization
Scenario 1: Initial synchronization Scenario 2: Synchronization is performed
and full copy are performed. again after the first synchronization, and
differential copy is performed.
a
b a a b Stored data
b
2 3
Snapshot c
1 2 a b c Newly copied data
Snapshot
1 c New data written to the host
a a
b b a a Internal signal flow
b b of the storage system
Source LUN Target LUN
c c
HyperClone pair
1 Create a HyperClone pair. Source LUN Target LUN
2 Create a snapshot for the

source LUN after 1 Create a snapshot for the source LUN
after a second synchronization.
synchronization is started.
3 Copy all data a and b to 2 Copy incremental data c to the target LUN.
the target LUN.
HyperClone Principles: Reverse Synchronization
Scenario 1: Full copy Scenario 2: Differential copy a b c Stored data
a a
a b c d Newly copied data
b b
c c
d New data written to the host
d d
2 1 2 1
Snapshot Snapshot Internal signal flow of the storage system
HyperClone pair
a a a a
b b b b
c c c c
d d d d
Source LUN Target LUN Source LUN Target LUN
1 Create a snapshot for the target 1 Create a snapshot for the target LUN after
LUN after the reverse the reverse synchronization is started.
synchronization is started.
2 Copy incremental data d to the source LUN.
2 Copy all data a, b, c, and d
to the source LUN.
Read/Write Principles of a Clone File System
Parent file File system Clone file Parent file File system Clone file
system snapshot system system snapshot system
Data status in the clone file system before data change Data status in the clone file system after data change
Application Scenarios: Data Backup and Restoration
Source LUN Target LUN
Time
point A a ⚫ Create HyperCopy.
⚫ Synchronize data on a source LUN to a target

Time LUN. In this case, the target LUN stores the data
a a
point B on the source LUN at time point B.
Time
point C a ⚫ Data on the source LUN is lost.
Time
⚫ Reversely synchronize data on the target LUN
point D a a to the source LUN. In this case, the source
LUN is restored to the status at time point B.
Application Scenarios: Data Analysis and Reproduction
Data analysis
Source Data analysis The data analysis service uses data on a
LUN host
target LUN to prevent the data analysis
service and production service from
Reproducing n contending for resources of a source LUN
pieces of data Only for data
analysis and affecting performance.
Data reproduction
Target Target LUN Target LUN HyperClone can create multiple copies of the
LUN (1) (n-1) (n) same source LUN for multiple target LUNs.
Application Scenarios: Application Test and Data Analysis
Start
Check the license.
Create a protection
Create a clone pair group (with LUNs).
(with LUNs or file
systems).
Create a clone
consistency group
(with LUNs).
End
Contents
1. HyperSnap
2. HyperClone
3. HyperCDP
▫ LUN HyperCDP
▫ File System HyperCDP
4. HyperLock
HyperCDP (for Block)
⚫ HyperCDP creates high-density snapshots on a storage system to provide continuous data protection. A
HyperCDP object is similar to a common writable snapshot, which is a point-in-time consistent copy of
original data to which the user can roll back to, if and when it is needed. It contains a static image of
the source data at the data copy time point.
⚫ Main features
 It provides intensive and persistent data protection. A single LUN supports 60,000 HyperCDP objects. The
minimum interval is 3 seconds.
 It provides data protection at an interval of seconds, with zero impact on performance and small space
occupation.
 It supports scheduled tasks. You can specify HyperCDP schedules by day, week, month, or specific interval.
 It supports HyperCDP consistency groups.
 A HyperCDP object cannot be directly mapped to a host for read and write operations but can be mapped
to a host after being converted to a writable snapshot by creating a duplicate.
Working Principles of HyperCDP
⚫ Based on the lossless snapshot technology, HyperCDP has little impact on the performance of
source LUNs. Compared with writable snapshots, HyperCDP does not need to build LUNs, greatly
reducing memory overhead and providing stronger and continuous protection.
⚫ HyperCDP objects cannot be mapped to hosts directly. To read data from a HyperCDP object, you
can create a duplicate for it and map the duplicate to the host.
Creating a HyperCDP Object
⚫ Creating a HyperCDP object is to save the data status of the source LUN at the activation time.
ROW principle Host data access

L3->P3
Source LUN CDP
L1->P1 L4->P4 T1 - - F - - - - -
T0 A B C D E - - -
L0->P0 L2->P2 L2->P5
L0 L1 L2 L3 L4 L5 L6 L7
• When a host accesses the source LUN data, the mapping

table of the source LUN is accessed from the T1 layer. If
A B C D E F - - the T1 layer does not exist, the data at the T0 layer is
P0 P1 P2 P3 P4 P5 P6 P7 returned. For example, if the host accesses data of L2, F
is returned. If the host accesses data of L0, A is returned.
• The HyperCDP mapping table is accessed from the T0
layer. If the host accesses data of L2, C is returned.
Reading and Writing a HyperCDP Object
⚫ A HyperCDP object cannot be directly mapped to a host for read and write operations
but instead, should be mapped to a host after being converted to a writable snapshot.
Source LUN CDP0 CDP1
Create a
snapshot
duplicate.
◆ First mapping Map
Host SNAP0
Source LUN CDP0 CDP1

Reactivate the snapshot.
◆ Second mapping Map

SNAP0
Host
Rolling Back a HyperCDP Object (1)
⚫ HyperCDP rollback is a process of copying data from a HyperCDP object to the source
LUN. The source LUN is available immediately after the rollback is started (data on the
source LUN is the data on a HyperCDP object).
Read data from the source LUN during
HyperCDP rollback and perform read redirection.
Source LUN
1 1. The host reads L1.
T1 L0 2. If L1 is not rolled back and the host has not written

L1 B' 3 data to L1 after the rollback is started, the host reads
2 data from CDP1 (T0) instead.
3. Data B is returned to the host.
CDP1
T0 L0 A
L1 B
Rolling Back a HyperCDP Object (2)
Write the source LUN during HyperCDP rollback.
Source LUN Source LUN
1. The host
L0 A' writes data L0 A'
T1 C to L1. T1
L1 B' L1 C
2. Roll back L0.
CDP1 CDP1
T0 L0 A T0 L0 A
L1 B L1 B
Source LUN Source LUN
T1 L0 A 4. The rollback T1 L0 A
L1 C is complete. L1 C
3. Roll back L1. Skip
CDP1 CDP1 this step if the host
has written data to L1.
T0 L0 A T0 L0 A
L1 B L1 B
HyperCDP Consistency Group
⚫ Medium- and large-size databases' data, logs, and modification information are stored on
different LUNs. If data on one of these LUNs is unavailable, data on the other LUNs is also invalid.
The HyperCDP consistency group ensures the consistency of application data during restoration.
⚫ Similar to individual HyperCDP objects, you can create, delete, roll back, or stop rolling back a
HyperCDP consistency group as required. You can also create or rebuild duplicates for HyperCDP
consistency groups.
CDP1
CDP2
LUN1 LUN1 Snapshot1
LUN2 LUN2 Snapshot2
CDP1-1
Create a
protection group. CDP2-1
Create a duplicate
Create a HyperCDP
for a HyperCDP
consistency group.
consistency group.
HyperCDP Schedule
⚫ You can specify HyperCDP schedules by day, week, month, specific interval, or any combination of them. You can also specify
the quantity of HyperCDP objects that can be retained for each schedule.
⚫ You can add multiple LUNs and LUN consistency groups to a HyperCDP schedule, but you can add a LUN or a LUN
consistency group to only one HyperCDP schedule.
⚫ A HyperCDP schedule supports the minimum interval of 3 seconds and retention of up to 60,000 objects.
Storage Pool Capacity Threshold
⚫ Because HyperCDP supports a minimum interval of 3 seconds, a large amount of data protection capacity
may be required if new data is writing to the source LUN constantly. As a result, the space of the storage pool
may be used up, affecting host services.
⚫ You can set a threshold to the used capacity of the storage pool and the protection capacity, respectively. You
can enable or disable it by using automatic deletion of protection data.
Used capacity
thresholds
Automatic deletion
Protection
capacity
thresholds
Application Scenario: Quick Data Backup and Restoration
⚫ HyperCDP backup allows fast data
restoration in the following scenarios:
If a source LUN covered by continuous data protection
 Virus infection is damaged, the source LUN's data can be recovered to
any point in time preserved by HyperCDP objects.
 Misoperations
 Malicious tampering
 Data corruption caused by system Source LUN HyperCDP HyperCDP HyperCDP HyperCDP
breakdown
 Data corruption caused by application
bugs
 Data corruption caused by storage 01:00:00 02:00:00 03:00:00 04:00:00
system bugs Continuous data protection Data flow
Time flow
Start
1. Check the HyperCDP

feature availability. Check the HyperCDP license.
2. Create a HyperCDP Create a protection group. Create a LUN.

object or HyperCDP
consistency group. Create a HyperCDP Create a
consistency group. HyperCDP object.
Legend: Mandatory Optional

End
Contents
1. HyperSnap
2. HyperClone
3. HyperCDP
▫ LUN HyperCDP
▫ File System HyperCDP
4. HyperLock
HyperCDP (for File)
⚫ HyperCDP creates high-density snapshots on a storage system to provide continuous
data protection. Based on the lossless snapshot technology, HyperCDP has little impact
on the performance of source file systems. Compared with common writable snapshots,
HyperCDP does not need to build file system objects, which greatly reduces the memory
overhead, delivering more intensive and continuous protection.
⚫ OceanStor storage system supports HyperCDP schedules to meet customers' backup
requirements. HyperCDP objects cannot be directly mapped to hosts. To read data from
a HyperCDP object, you need to share the file system with the host so that the host can
read the source file system data at the point in time when the HyperCDP object was
created.
Creating HyperCDP Duplicates
• After HyperCDP is enabled, the system
HyperCDP object's Source file system's generates a HyperCDP object for the file
mapping table mapping table system. After the HyperCDP object is created,
it shares the mapping table of the source file
system. In addition, the creation of the
HyperCDP object does not affect the data read
or write performance of the source file system.
The HyperCDP object is read-only.
L0->P0 L1->P1 L2->P2 L3->P3 L4->P4
• The original data in the source file system is

ABCDE. A HyperCDP object is created and
Storage space shares the mapping table of the source file
A B C D E
system. In the figure, L0 to L4 are logical
Physical address P0 P1 P2 P3 P4 P5 P6 P7 P8
addresses, P0 to P8 are physical addresses,
and A to E are the data.
Data Writing to the Source File System
• After the HyperCDP object is created,
HyperCDP object's Source file system's
mapping table mapping table the host writes data F and G to L0
and L2 of the source file system.
Because the ROW mechanism is used,
the host directly applies for new
physical addresses P5 and P6 for data
writing, and the original addresses P0
L0->P0 L1->P1 L2->P2 L3->P3 L4->P4 L2->P5 L0->P6
and P2 are referenced by the
HyperCDP object.
• The HyperCDP object does not
change the write process of the
Storage space A B C D E F G
Physical address P0
source file system or cause any extra
P1 P2 P3 P4 P5 P6 P7 P8
overhead. This avoids any loss of
performance.
Read Principle of HyperCDP
⚫ You can configure the NFS or CIFS sharing service in a storage system to share file systems with
clients. A HyperCDP object is a periodic read-only snapshot of the file system. After a HyperCDP
object is created, client applications can access the specified share path (.snapshot directory) in
the file system to read the source file system data at the point in time when the HyperCDP object
was created.
⚫ The scheduled snapshot can be accessed in the following three modes:
 Enter the .snapshot directory through the shared root directory or dtree directory of the file system to
access the data of any snapshot.
 You can go to the .snapshot directory from any directory. All scheduled snapshots for the directory are
displayed.
 You can enter the specific path to go to the .snapshot directory from any directory and access the
snapshot. Access to a scheduled snapshot that does not protect the previous directory will fail.
Application Scenario - Data Mining and Test Scenario
⚫ Duplicates can be created for HyperCDP objects and used for data mining and testing, which will not affect service data.
 A HyperCDP object is generated for the data to be tested at 11:00 a.m.
 A duplicate is created for the HyperCDP object and is read and written by the test server. During the test, the source data and services
accessing the source data are not affected.
 1 hour later, the source data and duplicate data are changed based on the data at 11:00.
 After the test, users can create another duplicate for the HyperCDP object to obtain the data at 11:00 and use the duplicate for another test.
Applications server Data at 11:00 a.m. Data at 12:00 a.m.
Applications
I/Os
Creates a duplicate
for the snapshot.
Test I/Os
Test server Snapshot at 11:00 a.m. Tested data at Another snapshot
12:00 a.m. at 11:00 a.m.
11 12 Time
Start
1. Check the HyperCDP

feature availability. Check the HyperCDP license.
Create a file system.
2. Create a HyperCDP object

or HyperCDP schedule. Create a HyperCDP Create a
schedule. HyperCDP object.
Legend: Mandatory Optional

End
Contents
1. HyperSnap
2. HyperClone
3. HyperCDP
4. HyperLock
Overview
⚫ Write Once Read Many (WORM), also called HyperLock, protects the integrity, confidentiality, and
accessibility of data, meeting secure storage requirements.
⚫ A file protected by WORM enters the read-only state immediately after data is written to it. In read-only
state, the file can be read but cannot be deleted, modified, or renamed. The WORM feature can prevent data
from being tampered with, meeting data security requirements of enterprises and organizations.
⚫ A file system with the WORM feature (a WORM file system for short) can be configured only by the
administrator. File systems with the WORM feature are classified into regulatory compliance WORM (WORM-
C) and enterprise WORM (WORM-E) according to the administrator's permissions.
Mode Application Scenario

This mode applies to archive scenarios where data protection mechanisms are
WORM-C
implemented as required by laws and regulations.
WORM-E This mode is mainly used by enterprises to implement internal control.
WORM Compliance Clock
⚫ To prevent users from changing protection periods of files by changing the system time,
storage systems maintain a WORM compliance clock. WORM compliance clocks include
the global security compliance clock and WORM file system compliance clock.
Clock Type Function Description

When creating a WORM file system for the first time,
The storage system maintains a
the system administrator must initialize the global
Global security global security compliance clock
security compliance clock. The time of the global
compliance clock that serves as the clock source for
security compliance clock cannot be changed after
all WORM file systems.
initialization.
The system will automatically use the global security

Each WORM file system
WORM file compliance clock to initialize the WORM file system
maintains a compliance clock.
system compliance clock upon the creation of a WORM file
The protection periods of files are
compliance clock system. You do not need to manually initialize the
based on the compliance clock.
WORM file system compliance clock.
File Status
⚫ There are four file states in a WORM file system, as described in the following table.
Status Description
All newly created files are in the initial state. Files in the initial state can be read,
Initial
written, and modified by all users.
Files in the locked state cannot be modified, deleted, or renamed by all users. These
Locked
files can only be read and their properties can be viewed.
Files in the expired state can be deleted and read and their properties can be viewed.
Expired
However, these files cannot be modified or renamed.
Data can be added to the end of files in the appending state and these files cannot be
Appending
deleted, truncated, or renamed.
WORM Properties
⚫ After the WORM feature is configured for a file system, the file system has the WORM properties. The WORM
properties apply to files in the WORM file system. You can view the WORM properties to determine the lock
time and expiration time of a file. The following table lists the WORM properties of a file system.
Property Description
Mode The system supports only the regulatory compliance mode.
Minimum retention period supported by the WORM file system. The retention period of a file in the WORM file
Min. Protection Period
system cannot be smaller than the minimum retention period.
Maximum retention period supported by the WORM file system. The retention period of a file in the WORM
Max. Protection Period
file system cannot be larger than the maximum retention period.
Default retention period supported by the WORM file system. The retention period of a file in the WORM file
Default Protection Period
system is the default value of the parameter if you do not set a retention period for the file.
After the automatic lockout function is enabled, files in the WORM file system automatically enter the locked
Automatic Lockout
state a specific period of time after data or metadata in the files is modified.
Default waiting time for modified files to automatically enter the protection state. This parameter is valid only
Lockout Wait Time
when Automatic Lockout is enabled.
After this function is enabled, the system automatically deletes files whose protection periods have expired.
Automatic Deletion Note: Before enabling this function, ensure that files do not need protection and can be automatically deleted
by the system after they expire.
Working Principles
⚫ If a common file system is protected by the WORM feature, files in the file system can be read only within the
protection period. After WORM file systems are created, you need to map them to application servers using
the NFS or CIFS protocol.
Extend the protection period.

Manually or automatically lock upon expiration.
Reset the protection period.
Initial Locked Expired

The protection
period expires.
Manually set the appending state.

Manually set the locked state.
(Condition: The file size is 0.)
Appending Manually set the appending state.

(Condition: The file size is 0.)
• A license file grants the permission to use a specific value-
Start
added feature. Before configuring a value-added feature,
ensure that the license file of the feature is valid.
Check the license file. • When creating a WORM file system for the first time, you
needs to initialize the global WORM compliance clock of the
storage system.
Create a WORM file system.
• After a WORM file system is created, storage resources can be
shared as file directories.
Share and access a WORM • After creating a WORM file system, share it with clients. You
file system.
can store files that need to be protected in the WORM file
system to prevent data tampering.
Manually set files to enter the
• After a WORM file system is created, if the automatic lockout
protection period.
function is not enabled, you may need to manually set files in
the WORM file system to enter the locked state on the client.
End In addition, you can set the files to enter the appending state
from the locked state to add content to the file.
Summary
HyperSnap
HyperClone
Storage data protection Definitions, principles,
and configuration
technologies and applications processes
HyperCDP
HyperLock
Quiz
1. (True or false) A source LUN can form multiple HyperClone pairs with different target
LUNs. A target LUN can be added to only one HyperClone pair.
2. (Short-answer question) When can a host read or write the source LUN after a rollback
command is executed?
Quiz
3. (Single-answer question) Which of the following is not a state of a file in a

WORM file system?
A. Initial
B. Locked
C. Deleted
D. Expired
Recommendations

⚫ Popular tools
 HedEx Lite
intelligent world.


that
Storage System O&M Management
Foreword
⚫ As the cost of storage devices decreases, large-capacity storage devices

have been used by more and more enterprises to store data generated by
enterprise service application systems and IT systems, such as emails,
documents, service data, and data backup. Therefore, effective
management of storage devices is critical to the continuity and stability of
enterprise services.
Objectives
Upon completion of this course, you will understand:

 General O&M management process
 Common storage system O&M management tools
 Process and methods of typical storage system O&M scenarios
Contents
1. O&M Overview
2. O&M Tools
3. O&M Scenarios
What Is O&M?
⚫ O&M is essentially the operation and maintenance of networks, servers, and services in each
phase of their life cycles to achieve a consistent and acceptable status in terms of cost,
stability, and efficiency.
• Organizational structure
• Asset management • Skill
• Event management • Personnel allocation
• Energy efficiency • …
management
• … Maintenance Personnel
management organization
Management Resource • Maintenance tool

• Device maintenance
Platform • Knowledge base
• Device repair Technical • Maintenance platform
• Risk identification assurance configuration • …
• …
Standard
Standard Service
system process
• ITIL • Standard operation procedure
• Uptime O&M • Emergency response process
• ISO 2000 • Troubleshooting process
• … • …
How to Perform O&M
⚫ Technical layer: Streamline the O&M lifecycle of each product and identify the key measures of
each task.
⚫ Process layer (ITIL process management framework): Change, event, and problem management.
Before the device goes online O&M/Operation period Process management

Model O&M/ Event management
Requirement Design Implementation Online Offline
selection Operation
Problem management
Requirement analysis Configuration selection and architecture design Offline specifications
Change management
Service rollout and transfer-to-maintenance criteria
Configuration
Rollout and deployment specifications management
Operation
Monitoring and security compliance
Release management
Capacity management and expansion
Performance and experience optimization
Emergency and DR drill
Cost and efficiency optimization
PRR management (performance, reliability, and recoverability)
Event Management
Objective
A
 Restore services as soon as possible.
 Minimize the impact of emergencies on service running.
 Ensure that the service quality and availability meet the SLA requirements.
Definition
B  Emergency
◼ Any event that causes or may cause service interruption or service quality deterioration
◼ Hardware faults, software faults, and service request interruption
Task
C  Detection and recording
 Classification and online support
 Priority determining based on the impact and urgency
 Investigation and diagnosis
 Solution and recovery
 End
 Responsibilities, monitoring, tracking, and communication
Event Management Process
Transfer and supervision
Service request
Accident data Problem

Service desk Accident management Temporary measure management
process:
Computer Change requirement Change
• Identification and
operation management
recording Solution
Input: accident
Report
Network • Classification and initial Service level
Output: support SLA parameter management
solution&
Procedure temporary
• Investigation and
Report Capability
measures analysis
management
Other sources • Solution and recovery
of accidents
• Accident termination Report Availability
management
Configuration information
Configuration management database

(CMDB)
Problem Management
Objective
A
 Locate the root cause of the problem and take measures to eliminate known errors.
 Minimize the number of emergencies caused by IT infrastructure errors and minimize the negative impact of
problems. Prevent the recurrence of emergencies related to errors.
Definition
B  Problem: obtained from multiple emergencies with the same symptom or a major incident and indicates
that an error with unknown causes exists.
 Known errors: The root cause of a problem has been successfully located and a solution has been found.
Task
C  Problem control
 Known error control
 Proactive problem management
◼ Trend analysis
◼ Review of major issues
Problem Management Process
Accident Configuration Service level Capability Availability

management management management management management
Information
Matching information, emergency

Record
measures, and quick recovery solution Problem
Problem
management Information database
Change request (RFC) Post implementation review (PIR)
Change
management
Change Management
Objective
A
 Ensure that all changes are effectively controlled and handled through standardized means and processes, and that
approved changes are implemented with minimum risks, high efficiency, and high cost-effectiveness.
Definition
B  Change: An action that causes the status of one or more IT infrastructure CIs to change.
 Standard change (approved in advance)
 Request for Change (RFC)
 Forward Schedule of Changes (FSC)
 Change Advisory Board (CAB)
Task
C  Receive, record, approve, plan, test, implement, and review change requests.
 Provide the IT infrastructure change report.
 Propel CMDB modification.
Change Management Process
The input information includes:
 Change request
 Data information provided by the CMDB, especially information about the impact of changes
 Change implementation schedule
 Capability database provided by capability management and budget information provided by
the financial management process
Change
The output information includes: management
 Updated change implementation schedule
 Signals that trigger the start of configuration
management and release management
 Agenda, minutes, and action items of CAB Problem Configuration Release
 Change management report management management management
Accident
management
Configuration Management
Objective
A
 Measure the value of all IT assets and configuration items used in organizations and services.
 Provide accurate information about IT infrastructure configuration for other service management processes.
 Support the operation of accident management, problem management, change management and release management.
 Verify the correctness of the configuration records related to the IT infrastructure and correct the detected errors.
Definition
B  Identify and define configuration items.
 Plan, define, and manage the CMDB.
 Periodically verify the accuracy and integrity of the CMDB.
 Detailed report of IT assets
Task
C  Receive, record, approve, plan, test, implement, and review change requests.
 Provide the IT infrastructure change report.
 Propel CMDB modification.
Configuration Management Process
Change management Release management Configuration management
Report and audit

Change request
configuration information.
Categorization
Report
and plan
Release Update information.

Configuration Definitive
management software
Release and distribute the database library
Implementation hardware and software of Update the CMDB and DSL.
the new version. (CMDB) (DSL)
Comment
Check whether the CMDB is

End
updated.
Publication Management
Objective
A  Comprehensively assess changes to IT services and ensure that all aspects (including technical and non-technical
factors) of a release are considered.
Definition
B  Release
◼ Delta release
◼ Full release
◼ Package release
 Emergency release
 Release policy
Task
C  Release planning
 Design, development, and configuration release
 Release review
 Rollout plan
 Communication, preparation, and training
 Distribution and installation
Release Management Process
Design and development
Controlled test environment Production environment
environment
Release management
Building
Release design, First-time Distribution
Release Release Release test and Communication
building, and running
policy planning and configuration acceptance and training
configuration planning installation
release
Configuration management database (CMDB)

Definitive software library (DSL)
Contents
1. O&M Overview
2. O&M Tools
3. O&M Scenarios
Components of Huawei Enterprise Storage O&M Systems
Huawei technical
Customer data center DME
Storage resource control support
O&M administrator
Resource pooling and service Trouble ticket generation and remote
Server orchestration troubleshooting
eSight eService cloud platform

Multi-device management
Fabric O&M administrator
Fault monitoring and performance report
Remote
command center
eService
Storage DeviceM DeviceM DeviceM
Remote maintenance
anager anager anager
O&M administrator
Fault monitoring and fault upload
Huawei Huawei Huawei Authentication system
OceanStor OceanStor hybrid OceanStor scale-
Dorado 18000 flash storage out storage SmartKit Equipment archive system
Storage service tool
Service personnel
DeviceManager Trouble ticket handling system
Delivery, upgrade, and troubleshooting
Single-device management
O&M administrator Fault diagnosis expert system
Configuration and alarm
Introduction to DeviceManager
⚫ DeviceManager is the single-device management software designed by Huawei for easy configuration,
management, and maintenance of storage devices.
⚫ Main software functions include storage resource allocation, user management, data protection feature
management, device performance monitoring, and alarm management.
DeviceManager GUI
Introduction to SmartKit
1. Unified platform
The desktop tool management platform integrates O&M
tools for storage systems, servers, and cloud computing.
2. Scenario-based guidance
Tools specific to each O&M scenario
can be downloaded on demand.
3. Standardized operations
The wizard guides you through
operations based on scenarios in
an easy and intelligent manner.
SmartKit GUI
⚫ Standardized and process-based operations in various service scenarios, improving operation efficiency
Function
navigation bar
Function
Field management
Device
Scenario-
based entry
Introduction to DME Storage
ITSM
Automation platform Monitoring platform Report platform
Open northbound API
Planning Construction O&M Optimization
SLA-based pooling and Flexible Automatic identification Multi-dimensional analysis and

integration automation modes and proactive prevention optimization suggestions
Intelligent planning and Orchestratable Automatic analysis and One-click change and
automatic pooling scheduled tasks quick locating autonomous driving
Gold Silver
Gold Silver Bronze
Open southbound API
Huawei OceanStor all-flash storage Huawei OceanStor scale-out storage FC switch Third-party storage
DME Storage Functions and Features
Planning Construction O&M Optimization
Proactive problem identification Automatic problem resolution

SLA-based pooling Automatic provisioning
and integration - SLA change
- Full-process automation - Configuration compliance/
Data protection - Data flow (reclamation/migration)
- Workload evaluation - Auxiliary automation
- Capacity/Performance threshold - Best practice implementation
- QoS policy - Orchestratable scheduled tasks
- Capacity/Performance prediction
- Threshold Policy - Application template
Multi-dimensional analysis
- Disk service life prediction
- Scheduling policy - SLA/Cost/Capacity/Performance
Automatic protection - Disk health/Abnormal workload
- Application/Device/Project/Label
- Quota policy
- Copy creation Automatic problem analysis - Customized report/dashboard
- Protection policy - Automatic
- Copy visualization
masking/aggregation/association Autonomous driving
- Copy enabling
Fast fault locating - Policy-based autonomy
- End-to-end performance analysis - AI-based inference autonomy
- End-to-end topology
Infrastructure
User role management Multi-DC management Open API
management
Introduction to eSight
⚫ eSight provides multi-vendor device adaptation for unified network-wide device management,
component-based architecture for on-demand construction of enterprise O&M platforms, and
lightweight design and web client for lower system maintenance and upgrade costs.
Switch/Router/WLAN
management
PON management …
Architecture UC/Telepresence and

Platform + Transport device
intelligent vision eLTE management
components management
management
eSight
Server/Virtual
Equipment room
Storage management resource
facility management
management
eSight basic management platform
Logical Architecture of eSight
Administrator
WLAN management Storage capacity management Server stateless computing
Network SLA and QoS management Storage network analysis Server configuration and deployment
Value-added O&M
Network traffic analysis and management Service management Report management
… … …
Unified communications device Network device eLTE device
Telepresence and videoconferencing Virtual resource PON device

Device management
Terminal Server device Transport device
Intelligent vision Storage device Equipment room facility
… … …
Management platform eSight management platform
Managed objects: servers, storage devices, VMs, switches, routers, firewalls...
Introduction to eService
⚫ Based on Cloud-Native, Huawei eService cloud intelligent management platform uses big data analysis and AI
technologies to provide services such as automatic fault reporting, capacity and performance prediction, and disk
risk prediction, preventing potential risks and providing a basis for capacity planning.
O&M Complex management

Reactive response Service rollout delay
challenges for multiple devices
Technical
Rapid response Real-time analysis Intelligent prediction Asset platform
capabilities
Cloud platform capability building

Proactive problem handling Intelligent prediction and analysis Asset management
 Provides 24/7 remote monitoring  Intelligently identifies risky disks,  Collects device asset information
for all IT devices, and automatic ensuring data security. for future reference.
eService fault reporting and SR creation.  Predicts the capacity trend in the  Provides product EOX
 Provides proactive problem next 12 months to facilitate information and device
handling to reduce the procurement planning. maintenance and warranty
troubleshooting duration and  Predicts performance bottlenecks information.
improve the operation efficiency. according to performance
fluctuation rules.
System Architecture of eService
Customer network Huawei network
Intranet DMZ
FusionCloud
ManageOne OC/SC Email (optional channel)

Only fault monitoring
supported
DC 1 Customer Huawei
DC 2
email email
Cloud OS server server
FusionManager/FusionCompute/Fusion Internet
Access/FusionSphere OpenStack OM eService
eSight
cloud
Storage/ system
Network Server/
Hardware layer HTTPS
Network
Storage Fault monitoring
eService
Remote inspection Proxy HTTPS
client
Server Remote log collection server service
& FusoinCare
Capacity prediction (optional)
Customer Performance
maintenance exception analysis
host Disk prediction
Contents
1. O&M Overview
2. O&M Tools
3. O&M Scenarios
Maintenance Item Overview
03
02
Weekly maintenance items
01
Daily maintenance items
 Inspecting storage devices
Initial maintenance items
 Checking the equipment room
 Checking and handling alarms
environment
 Checking the internal environment
 Checking the SmartKit installation
of racks
 Checking the installation and
configuration of eService
 Checking the alarm policy
configuration
Quick Maintenance Process
Start
Check the indicator status The indicators on the front and rear panels of the storage devices indicate the device
and handle related exceptions. running status and help you discover and rectify common hardware faults.
By checking the system information and service status, you can learn about the basic
Check the service status
information, alarms, capacity trend, and performance of the storage system, as well as
and handle related exceptions.
the status of storage resource allocation and value-added features.
Check the system performance

You can query the real-time and historical performance statistics on DeviceManager.
and handle related exceptions.
If a fault occurs in the storage system, DeviceManager automatically determines the

Check and clear alarms.
severity of the fault and reports an alarm. Maintenance personnel should rectify the
fault in a timely manner to prevent service interruption or data loss.
Can the Yes

preceding exceptions
be handled?
No
Collect information and report
faults.
End
O&M Scenario 1: Inspection
Background
⚫ After the storage devices purchased by company E are
deployed, services are deployed and running properly. To
ensure the storage security of core devices in the service
system, engineer A in the IT department is responsible for
the inspection of storage devices. Help engineer A make an
inspection plan.
Inspection Method
Checking the
storage system
operating
environment
Manual Inspection using tools:

Checking the SmartKit inspection
Checking inspection service DeviceManager inspection
indicators running status
Checking the
device running
status
O&M Scenario 2: Performance Monitoring
Background
⚫ Company E's business has gone online. To learn about the
performance usage of storage devices, engineer B in the IT
department is responsible for monitoring the performance
of the storage devices. Help engineer B monitor the
performance of the storage devices.
Performance Monitoring Process
Start
Different service scenarios have special requirements

on performance monitoring parameters. You can set
performance monitoring parameters based on service Configure performance
requirements. Ensure that the performance monitoring monitoring parameters.
function is enabled.
Configure the threshold. When the threshold is Configure thresholds.

exceeded, check the alarm information.
Create a service-related indicator template so that you Create a metric chart.

can view the indicators in a timely manner.
After the dump function is enabled, newly-generated
performance monitoring files are automatically Configure performance
dumped to the specified application server through monitoring file dumping.
the SFTP or FTP protocol.
View the performance monitoring
data.
View the storage system View the storage system alarm
View the storage device performance information, information.
including real-time performance monitoring data, performance information.
View the storage system event
alarms, events, and power consumption.
information.
View the storage system power
consumption information.
End
Performance Monitoring Metrics
IOPS • For applications with an I/O size smaller than 64

KB, mainly focus on the IOPS.
• For applications with an I/O size greater than or

equal to 64 KB, mainly focus on the bandwidth.
Monitoring
metrics
Bandwidth Latency
Performance Metrics
1 Snapshot 1 Logical port 1 Heterogeneous iSCSI link
2 Front-end Ethernet port 2 Host 2 Heterogeneous FC link
3 LUN priority 3 Controller 3 Remote replication consistency group
4 Back-end SAS port 4 LUN 4 FC replication link
5 Front-end FC port 5 Storage pool 5 Remote replication
6 Front-end bond port 6 SmartQoS policy 6 System
7 Disk 7 LUN group 7 Host group
O&M Scenario 3: Parts Replacement
Background
⚫ After the storage system of enterprise E has been running
for a period, it reports a disk failure and disk replacement
is needed. After receiving the disk replacement request,
Huawei technical support engineer C is going to the
customer site to perform the replacement.
Parts Replacement Process
Start
Log in to SmartKit. Remove cables Use the tool to

from the perform a post-
Select the component. replacement check.
component to be
replaced and start Remove the No
the replacement. Successful Rectify the fault as
component. check prompted.
Use the tool to Yes
perform the pre-
Insert a new
replacement check. Use the tool to
component.
perform the
inspection.
No Successful
Rectify the fault as
check Connect cables.
prompted. No
Successful Rectify the fault as
Yes check prompted.
Confirm that the
Use the tool to replacement is
power off the complete on the
Yes
component. tool.
End
Replaceable Parts
FRU CRU
Controller Power module
Interface module BBU module
System subrack Fan module
Management module Disk module
Cable Expansion module
Assistant cooling module Optical module
Quorum server -
Data switch -
Spare Parts Query
⚫ Huawei Spare Parts Query allows users to export the spare part manual and quickly obtain spare part
information based on the storage product model, part number, and SN.
⚫ You can log in to https://support.huawei.com/enterprise/en/index.html to use the Spare Parts Query
tool.
Key Points for Disk Replacement
⚫ When handling a disk module, hold only its edge to prevent damage.
⚫ Remove and insert a disk module with even force. Excessive force may damage the
appearance of the disk module or cause faults.
⚫ To avoid damaging disk modules, wait at least 1 minute between removal and insertion.
⚫ To prevent data loss, replace only a disk module of which the alarm/location indicator is
steady yellow.
⚫ Complete the replacement within five minutes after removing a disk module. Otherwise, the
system heat dissipation is compromised.
⚫ Use SmartKit to replace a risky disk (not faulty).
⚫ Ensure that the new disk is inserted into the same slot as the replaced disk. Otherwise, the
system may work abnormally.
Disk Replacement Using SmartKit
⚫ Start SmartKit and select the parts replacement tool.
Disk Replacement Wizard
Quiz
1. (Choosing multiple options) Which of the following are common management
software? ( )
A. DeviceManager
B. eSight
C. SmartKit
D. eService
2. (True or false) SmartKit integrates various tools required for deploying,

maintaining, and upgrading IT devices, helping product users, and service and
maintenance engineers perform precise operations on these devices, simplifying
operations and improving work efficiency. ( )
Summary
O&M concepts
O&M overview
O&M handling
process
DeviceManager
Storage system
O&M management SmartKit
Common O&M
eSight
tools
Common O&M
DME Storage
scenarios
eService
Recommendations

 Enterprise business: http://enterprise.huawei.com/en/
⚫ Popular tools
 HedEx Lite
intelligent world.


that

HCIA-Storage V5.0 Training Material

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

HCIA-Storage V5.0 Training Material

Uploaded by

Copyright:

Available Formats

Storage Technology Trends

⚫ Data carries information during the transmission on networks. What is the

On completion of this course, you will be able to understand:

1. Data and Information

3. Development of Storage Technologies

4. Development Trend of Storage Products

Input Processing Output

Item Data Information

Data Data Data Data Data Data

1. Data and Information

3. Development of Storage Technologies

4. Development Trend of Storage Products

⚫ Disaster recovery (DR) solutions

⚫ Storage management software

Network switching devices NAS gateway

Internal DAS SAN

Tape and single

Data Data Data

1. Data and Information

3. Development of Storage Technologies

4. Development Trend of Storage Products

Disks in a server Early external storage Storage arrays

CPU CPU CPU

RAM RAM RAM

Disks Disks Disks

SAN NAS Clients Unified

⚫ Generally, scale-out storage scatters

Back-end storage network

storage devices into logical resources,

Protocol and interface

Open Storage device and storage node

Universal hardware platforms

1. Data and Information

3. Development of Storage Technologies

4. Development Trend of Storage Products

1950s–1960s 1960s–1970s 1970s–1980s 1980s-1990s 1990s-

Single-level cell Triple-level cell Quad-level cell

Multi-level cell 3D TLC: improves SCM: with the access

Current Storage Technology SCM

1. Data and Information

3. Development of Storage Technologies

4. Development Trend of Storage Products

Parallel interface Serial interface

100G NVMe-oF 400G

NVMe 1.0 NVMe 1.2 NVMe 1.3 NVMe 1.4

1. Data and Information

3. Development of Storage Technologies

4. Development Trend of Storage Products

Steam Age Electricity Age Information Age Intelligence Age

Industry Cloud Service extension

Service Big data Customization

Gov & Public Large

Unstable latency EB-level TCO reduction

Biomolecule Optical storage

Blu-ray storage Gold nanostructured glass

50 years >>> 600 years

Start out Grow Lead

D. Massive amounts of data

B. A storage system consists of storage hardware, software, and solutions.

D. File storage is used to store data of data applications.

C. High power consumption

Data and Information

Data Storage Storage Architecture

Storage Products Interface Protocols