Professional Documents
Culture Documents
Notes:
The VMware vSphere APIs for Array Integration (VAAI) enables VMware to offload operations that
should be handled at the storage systems level to the SVC.
Some number of common VMware functions such as provisioning virtual machines from templates,
extending Thin Provisioning virtual disks (VMDK), and storage vMotion were accomplished using
numerous inefficient and repetitive write operations to storage systems.
VAAI enables integration with storage systems (or in this case the SVC) to make use of higher
performance functionality such as the same SCSI write command to handle blocks of zeros or the
SCSI extended copy (XCOPY) command to minimize server I/O and reduce network bandwidth.
Uempty
VMware traffic
to/from its
datastore
SVC traffic
to/from its
backend storage
Notes:
The GUI real time performance statistics graphs can be used to illustrates the benefit of the VAAI
integration with the SVC.
The traffic shown from VMware to the SVC as 380 MBps before VAAI integration.
© Copyright IBM Corp. 2011, 2014 Unit 4. SVC host access 4-95
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
VMware traffic
offloaded to
SVC backend
Notes:
The visual shows that the VMware to SVC traffic has been removed.
Since the SVC is the storage system from a host (VMware) perspective, the SVC’s support for VAAI
means the back-end storage systems managed by the SVC do not have to implement VAAI.
Uempty
Checkpoint (1 of 2)
1. True or False: Zoning is used to control the number of paths
between host servers and the SVC.
Notes:
Write your answers here:
© Copyright IBM Corp. 2011, 2014 Unit 4. SVC host access 4-97
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
Checkpoint (2 of 2)
3. True or False: A multipath driver is needed when multiple
paths exists between a host server and the SVC cluster.
Notes:
Write your answers here:
Uempty
Unit summary
Having completed this unit, you should be able to:
• Summarize the SVC volume access infrastructure
• Differentiate between a volume’s caching I/O group and accessible I/O
groups
• Identify the number of instances a LUN is reported to a host OS for a fabric
configuration
• Implement subsystem device driver (SDD) commands to monitor device
path configuration
• Verify zoned ports between a host to the SVC and between a storage
system to the SVC using either the CLI commands or the Fibre Channel
interface
• Perform nondisruptive volume movement from one caching I/O group to
another
• Configure SVC Ethernet ports to enable iSCSI initiator connectivity
Notes:
© Copyright IBM Corp. 2011, 2014 Unit 4. SVC host access 4-99
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
Uempty
Unit 5. SVC enhanced features
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-1
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
Uempty
Unit objectives
• After completing this unit, you should be able to:
• Recognize Easy Tier settings and statuses at the storage pool and
volume levels
• Differentiate among fully allocated, Thin-Provisioned, and compressed
volumes in terms of storage capacity allocation and consumption
• Create thin-provisioned volumes and monitor volume capacity utilization
of autoexpand volumes
• Use Volume Mirroring to convert a volume to Thin-Provisioned or
compressed to reclaim allocated but unused capacity
• Itemize SVC hardware resources required for Real-time Compression
• Use the Comprestimator utility to identify suitable candidate volumes for
compression
• Analyze Easy Tier Storage Tier Advisor Tool (STAT) reports for volume
heat distribution and capacity recommendations
Notes:
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-3
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
• Thin Provisioning
• Volume Mirroring
• Real-time Compression
• Comprestimator utility
Notes:
This topic introduces the v7.3 enhanced features beginning with Easy Tier v3.
Uempty
Notes:
IBM Storwize Family Software v7.3 improves Easy Tier capabilities to allow for up to three tiers of
storage. This table lists the supported combination within a storage pool tier. Tiers are identified as
Flash/SSD which represents solid state drives, flash drives, or flash storage that is being
virtualized. Hard disk drives are separated into two tiers: Enterprise 15 K and 10K SAS drives are
both classified as ENT, and NL is the Nearline 7.2 RPM drives. An example, you could have
Flash/SSD, ENT, and NL as Tier 1 (1), or you could have ENT and NL (4), or a single tier using any
of the three tiers listed. SSD and Flash drives are always Tier 0.
MDisks in the SVC will always be displayed as Enterprise (ENT) unless you are using the 2145-24F
expansion enclosure that supports only SSD Flash drives. Otherwise, you must change the
description of the tiers on the SVC so it can identify them correctly.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-5
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
Notes:
With v7.3, the Easy Tier definitions remain the same as before. The concept of Easy Tier is to move
data up and down based on performance between three tiers (performed in the same manner as
the movement between two tiers), and the heat of extents. Easy Tier is transparent to the user once
you place multi-tiers in a pool it is automatically enabled.
When creating new volumes, they are placed by default on the Enterprise or middle Tier 1. If Tier 1
has reached its capacity, then it will used the next lowest tier which would be Tier 2. If all tiers are
full, only then will it allocate extents from Tier 0. Easy Tier will then start migrating those extents (hot
or cold) based on the workload.
For best practices, it is recommended to keep some free extents with in a pool in order for Easy Tier
to function. This will allow Easy Tier to move the extents between tiers as well as move extents with
in the same tier to load-balance the MDisks with in that tier, without delays or performance impact.
Easy Tier will work using only one extent; however, it will not work as efficient.
Free extents can be estimated based on one extent times the number of MDisks in a storage pool
plus 16.
Uempty
Vol2 Vol3
Vol0 Workload
Vol1 balanced across all 3 MDisks
Figure 5-5. Easy Tier v3: Automated Storage Pool Balancing SNV13.0
Notes:
SVC also introduces support for Automated Storage Pool Balancing, an added feature within the
7.3 code which, while associated with Easy Tier, operates independently of Easy Tier and does not
require an Easy Tier license. This eliminates the need of using a script to restrip extents across the
storage pool. Automated Storage Pool Balancing assesses the extents that are written in a pool and
balances them automatically across all MDisks within the pool. In this case, you can have a single
tier pool or mix different drive type and capacity that you have MDisks on in the same pool.
Automated Storage Pool Balancing will identify the stanzas of those internal drives by RAID levels,
drive types, sizes, and the number of drives in each pool as well as its IOPs capabilities. This is only
a performance rebalance – not an extent rebalance. It will then balance those extents across all the
MDisks in the pool equally.
Automated Storage Pool Balancing can work in conjunction with Easy Tier when multiple classes of
disks exist in a single pool. The process will automatically balance existing data when new MDisks
are added into an existing pool even if the pool only contains a single type of drive.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-7
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
Notes:
Automated Storage Pool Balancing uses XML files that are embedded in the v7.3 code. The XML
files uses stanzas to records the characteristics of the drives by RAID levels that are built, the width
of the array, drive types, and sizes used in the array, and so on, to determine MDisk thresholds.
Automated Storage Pool Balancing can be disabled on the pool.
Uempty
Hybrid
storage pool
Hot extents
Cold extents migrate up
migrate down
Extent size
Size: 11 GB 1024 MB
HDD MDisks
Notes:
Easy Tier supports two modes of operation:
• Automatic data placement mode: Optimizes SSD/flash usage
• Evaluation mode: Assesses potential workload benefits if SSD/flash investment
Easy Tier automatic data placement mode is enabled by default when a multi-tiered storage pool is
detected. All the volumes of the multi-tiered storage pool are automatically managed by Easy Tier.
Active extents of these volumes are eligible to be migrated dynamically to SSD-based MDisks to
achieve better performance. This sub-volume extent movement is transparent to host servers and
applications.
It is possible to disable a given volume of a multi-tiered pool from Easy Tier automatic data
placement by changing the volume's Easy Tier setting to off. This action causes the volume's I/O
statistics to be measured by Easy Tier but no extent movement occurs.
A statistic summary file or ‘heat’ file generated by Easy Tier can be offloaded for input to the IBM
Storage Tier Advisor Tool (STAT). This tool produces reports on the amount of extents moved to
SSD-based MDisks and predictions of performance improvements that could be gained if more
SSD capacity is available.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-9
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
Uempty
Hot extents
Cold extents migrate up
migrate down
HDD MDisks
Notes:
Prior to any SSD acquisition, the Easy Tier setting for a single tiered storage pool can be changed
to on. This causes Easy Tier to run in evaluation mode for the pool.
I/O activity of the volumes in the pool are measured and an I/O statistics file is generated. This file,
also known as the ‘heat’ file, can be offloaded from the SVC cluster and analyzed by the IBM
Storage Tier Advisory Tool (STAT). This analysis provides an understanding of the benefits for the
workload if SSD-based MDisks were to be added to the pool.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-11
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
Notes:
The hot and cold temperature of an extent is dependent upon the measurements of random, small
data transfer I/O operations.
Large sequential transfers are not considered as these tend to perform equally well with
HDD-based MDisks. Thus, Easy Tier only considers extents with I/Os of up to 64 K as migration
candidates.
Uempty
Notes:
Easy Tier is controlled by a combination of settings at both the storage pool and volume levels.
A storage pool Easy Tier setting has a default value of auto. It can be changed to on or off with the
CLI. A storage pool’s Easy Tier status can be either inactive or active - this value is a result of its
setting value and if the pool is single or multi-tiered.
The GUI displays the pool’s Easy Tier status of inactive or active. The CLI also displays this status
information.
In addition, the CLI displays the pool’s Easy Tier setting of auto/on/off - because the way to change
this setting is through the CLI. The GUI does not provide an interface to change the Easy Tier
setting of the pool.
The default pool setting of auto enables Easy Tier to be automatically activated when the pool is
detected to have more than one tiers of storage technology. Since the Easy Tier setting for the
DS3K_SASpool is auto and its Easy Tier status is inactive, it must only contain one tier of storage at
this point in time.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-13
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
IBM_2145:RED_SVC:REDadmin>lsvdisk WISK1
id 3
name WISK1
Volume Copy settings only altered IO_group_id 0
IO_group_name io_grp0
in CLI status online
mdisk_grp_id 0
mdisk_grp_name DS3K_SASpool
Easy Tier On/Off capacity 5.00GB
. . .
setting copy_id 0
. . .
easy_tier on
Easy Tier Inactive/Active/Measured easy_tier_status inactive
tier generic_ssd
status tier_capacity 0.00MB
tier generic_hdd
tier_capacity 5.00GB
Note: Easy Tier is implemented at the volume copy level
Notes:
To access the Volume Details, a user would right-click the volume followed by clicking Properties.
At the volume level, the Easy Tier setting has a default value of on. It can be changed to off and on
with the CLI. A volume’s Easy Tier status can be inactive, active, or measured - this value is
dictated by the volume level Easy Tier setting value along with the Easy Tier status of the pool.
The GUI displays the volume’s Easy Tier status of inactive, active, or measured. The CLI also
displays this status information.
In addition, the CLI displays the volume Easy Tier setting of on/off - because the way to change this
setting is through the CLI. The GUI does not provide an interface to change the Easy Tier setting of
the volume.
The default volume setting of on allows a volume to be automatically managed by Easy Tier once
its pool becomes Easy Tier active. The default settings for the pool and volume enable automated
storage tiering to be implemented without manual intervention.
Since the Easy Tier setting for the WISK1 volume is on and its Easy Tier status is inactive, the Easy
Tier status of its pool must be inactive at this point in time.
Uempty
2
3
1
Figure 5-12. Support for SVC internal solid state drives SNV13.0
Notes:
The SVC internal SSDs can reside within the SVC nodes or expansion enclosure. In the CG8 or
CF8 models that are grouped by the I/O group of the owning nodes. These SSDs are displayed by
the GUI Pools > Internal Storage view.
The installed SSD drives of the SVC cluster are listed by drive class. Each drive class represents a
unique SSD drive capacity installed (a variety of SSD drive capacities are supported). The Drive
Class Filter buttons allow a given drive class to be displayed while the default displays of all drives
installed. In this example, the 136.2 GB io_grp0 push button enables the display of the eight drives
installed in NODE1 and NODE2 (both nodes being members of io_grp0). Each drive has an
assigned drive object ID.
Internal SSDs are configured as RAID arrays, also known as array mode MDisks, and placed into
pools. Click the Configure Storage button to start the SSD RAID configuration.
If the use attribute of a drive is unused, the GUI automatically presents a pop-up window to prepare
it for RAID array configuration. Click Yes to change the use attribute of the drives.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-15
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
Notes:
The GUI generates the appropriate number of chdrive commands to change the use attribute of
each discovered drive from unused to candidate.
Internal drives must have a use attribute of candidate before an array mode MDisk can be created.
Click Configure Storage again to configure RAID arrays (or array MDisks).
Uempty
Figure 5-14. RAID array GUI presets: SVC internal SSDs SNV13.0
Notes:
Under the previous software code, the SVC 2145-CG8 and F8 supports only have the option to
select RAID levels 0, 1 and 10. However, with the release of the v7.3 code, the SVC GUI now offer
five presets options in which to configure internal or expansion SSDs into RAID-managed disk
arrays. IBM 2145-DH8 supports RAID levels 0, 1, 10 plus RAID levels 5 and 6.
The SSDs within the CG8 or CF8 nodes are internal to the SVC nodes, the volumes created should
be owned by the same I/O group or nodes that contain the SSDs (to minimize routing between I/O
groups).
With DH8 ability to attach up to 48 SSDs using 12 Gb SAS cables instead of having four smaller
SSDs inside the node, provides ability to create RAID array across both expansions attached to
each node in the I/O group.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-17
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
Array
MDisk
v7.1 GUI
6
© Copyright IBM Corporation 2011, 2014
Figure 5-15. Configure SVC internal storage: Easy Tier option SNV13.0
Notes:
This example uses the RAID Presets supported on the CG8 and CF8 nodes. Easy Tier preset is
selected to create an array MDisk then adds it to an existing storage pool; causing Easy Tier to be
automatically activated for the hybrid pool.
The Easy Tier RAID 1 drive configuration preset has been selected. Even though the system has
found or discovered eight drives, two drives have been specified as the number of drives to
provision for this example. Of the remaining drives, one per node will be defined as a spare drive.
Uempty
DS3K_SASpool
SSD
7 array MDisk MDisk MDisk
MDisk
10
© Copyright IBM Corporation 2011, 2014
Notes:
This array MDisk can be placed into an existing pool or a new pool. The DS3K_SASpool has been
chosen since the goal is to change this pool’s composition to a hybrid pool.
The mkarray command generated by the GUI creates a RAID level 1 array using two drive IDs
(one SSD from each node), and places the array MDisk in the selected storage pool.
The use attribute of each of these two drives has been updated to member as they are now
members of a RAID array.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-19
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
2
3
4
© Copyright IBM Corporation 2011, 2014
Notes:
One of the SSD drives with the use attribute of candidate is being configured as a spare. If an SSD
drive in the array MDisk fails, automatic RAID rebuild with the spare drive will restore the RAID
array’s redundancy.
In this example, a spare drive has been designated for each node.
Uempty
Figure 5-18. SSD MDisk added to pool: Easy Tier active SNV13.0
Notes:
Go to Pools > MDisks by Pools view and examine the pool icons associated with the three
storage pools. Focus on the DS3K_SASpool icon. An Easy Tier active flag (disk with stripes) has
been added to the pool icon to indicate that Easy Tier has been activated for the pool.
MDisk ID 17, with an SVC assigned name of mdisk0, is the array mode MDisk that has been
added to the pool. Note its Tier column value indicates the MDisk’s technology tier is solid state
drive.
The new MDisk is still being initialized as a RAID array (one array initialization task in the Running
Tasks status pod). Application I/O to the MDisk is supported during RAID array initialization.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-21
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
Notes:
The Pools > Volumes by Pool view of the DS3K_SASpool provides a close-up of the Easy Tier
active pool icon and confirms that Easy Tier has been automatically activated for this pool. Do note
that the caching or owning I/O group for the volumes of the pool should align with the SSD-based
MDisks when internal SSD drives are used.
Examine the details of the WISK1 volume. Its Easy Tier status is now active. At this point, all of the
extents of the volume are still sourced from the HDD technology tier but they are now subject to
Easy Tier measurement. It takes around 16 - 24 hours of data collection and assessment for Easy
Tier to determine the ‘temperature’ of the extents and implement its migration plan.
To summarize; the presence of the array MDisk in this pool has caused Easy Tier to become active
automatically (recall the setting for the pool defaults to auto). Because Easy Tier is active for the
pool, the volume Easy Tier status became active (recall its default setting for the volume is on).
These Easy Tier default settings enable automated storage tiering optimization without additional
administrative effort.
Uempty
2 4
3
5
© Copyright IBM Corporation 2011, 2014
Notes:
MDisks of external storage systems can be backed by either HDD or SSD technology (SSD also
includes flash). By default, all external MDisks are assigned the HDD tier since there is no interface
for the SVC to discern its storage technology. SVC does provide an interface to identify the correct
tier for these MDisks.
This example illustrates an unmanaged SSD-based MDisk (object ID 19) surfaced from a Storwize
V7000 system. It has a default tier of Hard Disk Drive. Right-click this MDisk entry and click Select
Tier to alter its tier assignment.
The chmdisk command generated for MDisk 19 changes its storage tier to generic_ssd tier.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-23
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
1
2
4
© Copyright IBM Corporation 2011, 2014
Notes:
The unmanaged MDisk ID 19 with the updated solid state drive tier assignment is then added, with
the Add to Pool option, to the SWV7K_SASpool.
Uempty
SWV7K_SASpool
SSD
MDisk MDisk MDisk MDisk
MDisk
DS3K_SASpool
SSD
array MDisk MDisk MDisk
MDisk
Notes:
Easy Tier is now active for the SWV7K_SASpool because the newly added MDisk changed this
pool to a hybrid with two tiers of storage. The Tier column shows a mixture of both HDD and
SSD-based technology for the MDisks of this pool.
This environment now has two pools with Easy Tier active. The SWV7K_SASpool contains a
managed mode SSD-based MDisk sourced from external storage. The DS3K_SASpool contains an
array mode SSD-based MDisk sourced from SVC internal SSDs. Easy Tier provides the same
functionality regardless where the SSD-based MDisk is sourced.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-25
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
Notes:
To obtain a quick review of the Easy Tier status of a list of volumes, use the Volumes > Volumes
view and add the Easy Tier Status column to the display. If desired, the search box allows a more
focused list to be displayed.
More examples of Easy Tier behavior are included in the ensuing topics. This unit finishes with the
Easy Tier heat map file and STAT (Storage Tier Advisor Tool) reporting.
Uempty
• Thin Provisioning
• Volume Mirroring
• Real-time Compression
• Comprestimator utility
Notes:
This topic discusses the Thin Provisioning concept.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-27
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
Notes:
The SVC Thin Provisioning function extends storage utilization efficiency and economy to all SVC
supported storage systems. A volume can be created either as fully allocated or Thin-Provisioned.
A thin-provisioned volume is created with two capacities - virtual and real. The virtual capacity is the
capacity seen by host servers as well as SVC services such as FlashCopy and Metro/Global
Mirroring. The real capacity defines the amount of capacity actually allocated from disk systems
represented by storage pools. The real capacity can be increase either manually (command line
default) or automatically (GUI default).
For fully allocated volumes, its entire capacity is acquired at creation and the entire logical block
address (LBA) range is allocated at the same time. For thin-provisioned volumes, only the real
capacity is acquired at its creation. Space within the allocated real capacity is allocated in
grain-sized increments corresponding to an LBA range dictated on write activity.
Uempty
2 GB
Notes:
Since only the virtual capacity is presented to attaching hosts, the implementation is totally
transparent to attaching hosts. Hosts have no awareness hence no need to discern between fully
allocated versus thin-provisioned volumes.
The SVC implementation of Thin Provisioning is also totally transparent to back-end storage
systems. As a standard feature of the SVC, thin-provisioned volumes might reside in any storage
pool representing any attaching storage system. In addition, thin-provisioned volumes can reside in
the same pool as fully allocated volumes.
In addition to optimize storage capacity utilization effectiveness to improve returns on investment,
Thin Provisioning also expands opportunities to exploit additional data protection options. For
example, enabling more frequent recovery points of business data to be taken (point in time copies)
without a commensurate increase in storage capacity consumption.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-29
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
Capacity
(size) Volume
Virtual Capacity: 5 GB
Extents
Real Capacity: 2 GB
Space allocated in grain size increments:
• Default grain size = 256K
• Grain size can be 32/64/128/256K
LBA2014
Write/Read
Activities: (Grain=256K)
LBA0-511 LBA1536-2047
LBA463
Notes:
When a thin-provisioned volume is created, the real capacity acquired is controlled by the rsize
(real size) parameter. This capacity can be manually increased unless the autoexpand parameter
is set on (CLI default is off).
Space within the real capacity is allocated in grain-sized increments each representing an LBA
range. The default grain size is 256 K - which represents 512 blocks of 512 bytes each. A write
operation to an LBA range not previously allocated causes a new grain sized space to be allocated.
A directory is used to map or track allocated space to a corresponding LBA range based on write
activities. This directory of metadata and user data shares the real capacity allotment of the volume.
When write activity exceeds its real capacity, the volume goes offline (if the automatically expand
setting is off) and application I/Os would fail. Once the real capacity is expanded, the volume
becomes online automatically. To avoid exhausting the real capacity, alerts can be sent to
administrators to increase real capacity and/or change the volume attribute to autoexpand so that
real capacity is increased automatically.
Uempty
Write/Read Contingency
Activities: capacity
(rsize=2GB)
(Grain=64K)
LBA0-127 LBA256-383 LBA128-255
LBA200
Notes:
The autoexpand keyword, which is the GUI default, enables the real capacity of a thin-provisioned
volume to automatically increase as needed. This automatically expand attribute can be requested
either at volume creation or subsequently.
With the autoexpand attribute set on, the rsize specified value serves as a reserved capacity that is
maintained as write activities occur. This reserved or contingency capacity protects the volume from
going offline in the event the storage pool runs out of extents. The contingency capacity diminishes
to zero when the real capacity reaches the total capacity of the volume.
The combination of autoexpand and the existence of the metadata directory might cause the real
capacity of the volume to become greater than the total capacity seen by host servers and other
SVC services. A thin-provisioned volume can be converted to fully allocated using the SVC Volume
Mirroring function.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-31
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
Lower Cache
Forwarding
Volume’s
RAID
extents
Forwarding
SCSI Initiator
Notes:
When a thin-provisioned volume is initially created, a small amount of the volume’s real capacity is
used for the volume’s metadata directory. A write I/O to an LBA that have not previously been
written to will cause a grain-sized amount of space to be marked as used within the volume’s
allocated real capacity; and its metadata directory updated.
For example, if the volume default grain size is 256, then 256 K within the allocated real capacity is
marked as used, for the 512 blocks of 512 bytes each, spanning the LBA range in response to this
write I/O request.
If a subsequent write I/O request is to an LBA within the previously allocated 256 K, the I/O
proceeds as usual since its requested location is within the prior allocated 256 K.
If a subsequent write I/O request is to an LBA outside the range of a previously allocated 256 K,
then another 256 K within the allocated real capacity is used.
All three of these write examples consult and might update the metadata directory. Read requests
also need to consult the same directory. Consequently, the volume’s directory is highly likely to be
SVC cache-resident while I/Os are active on the volume.
Uempty
3
5
4 Optional
Notes:
Choose the Thin-Provision preset of the GUI to create a thin-provisioned volume.
The capacity value is the virtual capacity of the volume; it is the size presented to hosts and other
SVC services such as FlashCopy and Metro/Global Mirror.
Review the Summary statement, for the 100 GB APPVOLA volume requested to reside in the
DS3K_SASpool, only 2 GB of real capacity would be allocated. This 2 GB default value is controlled
by the Thin-Provision preset. Click the Advanced button to view and/or override preset attributes.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-33
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
Thin-Provisioned
defaults and
options
Notes:
The Thin-Provisioned option of the Capacity Management tab displays its preset defaults. The
edit box for Real has a default of 2% of virtual size, which explains the 2 GB value for the 100 GB
volume allocation request.
The Thin-Provision preset also sets the automatically expand attribute to on, a warning threshold of
80% (of volume size), and a grain size of 256 K. These attributes can be overridden from this pane;
but as a general best practice, take the preset defaults.
Recall the DS3K_SASpool is Easy Tier active and contains an internal SSD-based MDisk from
io_grp0. Due to the dependency of this MDisk and hence to pool to this I/O group (I/O group
preference or allegiance), the GUI automatically assigns the volume ownership to its preferred I/O
group - io_grp0.
Uempty
Thin-Provisioned
volume created;
object ID 11
Notes:
Review the generated mkvdisk command. The -autoexpand, -grainsize 256, -rsize 2%,
and -warning 80% keywords are associated with volumes created with the Thin-Provision preset.
The APPVOLA volume is created in the DS3K_SASpool with an object ID of 11 and I/O group 0
assigned as its caching I/O group.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-35
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
thin-provisioned
volume
APPVOL1 APPVOLA
DS3K_SASpool
SWV7K_SASpool
Notes:
The ensuing pages illustrate the usage of a thin-provisioned volume assigned to host object
REDWIN1. To compare and contrast, a fully allocated volume is also assigned to the same host.
The two volumes are:
• APPVOL1 is a 100 GB fully allocated volume in the SWV7K_SASpool.
• APPVOLA is a 100 GB thin-provisioned volume in the DS3K_SASpool.
For later reference, note that volume APPVOL1 has an ID of 9 with a UID that ends with x’000D’;
volume APPVOLA has an ID of 11 with a UID that ends with x’0013’.
Uempty
APPVOLA
Notes:
The capacity bar of the volume details pane makes it plainly obvious that the 100 GB APPVOLA
volume is a thin-provisioned volume. Only a tiny amount, 2 GB of real capacity, is allocated; within
which 768 KB is used: For the volume metadata directory.
The Member Disks tab displays the extent distribution of this volume as well as the actual number
of extents consumed on each MDisk. Hmmm five extents of 512 MB work out to be more than 2 GB.
What is going on?
Recall that the automatically expand attribute defined for this volume. The real capacity (rsize)
value of 2 GB (or four extents) serves as a contingency buffer maintained for this volume. Because
768 KB of real capacity has been used, the volume was automatically expanded by one extent to
maintain the 2 GB buffer.
The volume’s Easy Tier status is active since its pool’s Easy Tier status is active.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-37
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
write
activity
Host sees
100 GB volume
© Copyright IBM Corporation 2011, 2014
Notes:
From the host perspective, the thin-provisioned volume appears as a fully allocated volume.
Based on the SDDDSM datapath query device command output, Disk1’s serial number
matches the UID of the SVC volume APPVOL1 and Disk2’s serial number matches the UID of
volume APPVOLA.
Both volumes are presented to the as volumes with 100 GB capacity. A folder of data has been
written to driver letter E: or APPVOLA. Examine the capacity and free space data tracked by
Windows for APPVOLA - it has close to 77 GB of free space.
Uempty
APPVOL1
Notes:
Contrast the thin-provisioned volume to a fully allocated APPVOL1 volume. Again the volume’s
capacity bar makes it plainly obvious; all 100 GB of the volume capacity (or 200 extents of 512 MB
each) have been allocated at volume creation.
Relative to Easy Tier, this volume has an Easy Tier status of active and 22 of its extents have been
migrated to the SSD-based MDisk of the pool.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-39
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
11
APPVOLA
Notes:
Examine the details of the APPVOLA volume after host write activity. The volume capacity bar
highlights the increased consumption of real capacity.
The used capacity value of 23.15 GB is fairly consistent as the complement value of the 77 GB free
space shown on the host.
Note the real capacity value of 25.16 GB? Why is it about 2 GB larger than the used capacity?
To minimize capacity consumption, the SVC 2145-CF8 and 2145-CG8 node models provide a zero
detect function. When host writes result in a grain sized string of zeros, no actual capacity is
allocated.
Uempty
more
write
activity
Notes:
The host has written significantly more data to the thin-provisioned volume. Notice the% Free
column for the APPVOLA(E): drive is only 17%.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-41
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
11
Notes:
Again, the capacity bar for the APPVOLA volume makes it plainly obvious that the space warning
threshold has been exceeded. The used and real capacity values both have exceeded the 80 GB
(or 80% of volume capacity) threshold.
Uempty
11
APPVOLA
Resolution options:
Ź Expand volume size
Ź Convert to fully allocated
© Copyright IBM Corporation 2011, 2014
Notes:
When a warning threshold is exceeded (for a volume or a pool), the SVC generates event
notifications. For each event, the notifications include SNMP trap, email notification, and an entry in
the SVC event log.
To see the event log threshold warning entry, go to Monitoring > Events and select the Unfixed
Messages and Alerts view of the log. An alert has been generated for the APPVOLA volume.
Right-click the entry and select Properties for a more readable and detailed description of the
event. A thin-provisioned volume copy space warning has occurred.
Assuming that the application has estimated the volume size correctly, a fully provisioned volume
might be more appropriate. Otherwise, the application might want to reassess the volume virtual
size.
A thin-provisioned volume can be converted to a fully allocated volume by using the SVC Volume
Mirroring function.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-43
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
11
Host deletes
do not release APPVOLA
previously
allocated extents
Notes:
Data or file deletion is managed at the host or OS level. In Windows, a file deletion is just an update
of allocation tables associated with the Windows drive to release allocated space. It is an activity
not known to external storage systems including the SVC.
Consequently, the real capacity utilization of the SVC volume is not changed nor released even
though the host system indicates more free space for the drive.
Uempty
Notes:
Examine the output of the lsvdisk 11 command. Recall that SVC objects can be referred to by ID.
The command lsvdisk APPVOLA provides the same information.
Note the overallocation value of 117. Overallocation at the volume level is a useful indicator of the
convergence of real and virtual capacity. As this value gets close to 100, it means the real capacity
is growing closer to virtual capacity. At 80% of the volume’s warning threshold value (or 80 GB of a
100 GB volume’s real capacity has been consumed), the overallocation value would be at 125.
As the volume’s overallocation value gets close to 100 (or even at 125), a key question arises -
should this volume have been allocated as a thin-provisioned volume? A reassessment if this
volume should be Thin-Provisioned might be in order; perhaps it would be a good idea to convert
the volume to fully allocated.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-45
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
List volumes that have consumed 60% or more of real capacity of volume
– filter on “overallocation less or equal to 166” (where 166=((100/60) * 100))
IBM_2145:RED_SVC:REDadmin>lssevdiskcopy -delim , -filtervalue
"overallocation<=166"
vdisk_id,vdisk_name,copy_id,mdisk_grp_id,mdisk_grp_name,capacity,use
d_capacity,real_capacity,free_capacity,overallocation,autoexpand,war
ning,grainsize,se_copy,compressed_copy,uncompressed_used_capacity
6,SLIM2,0,2,SWV7K_SASpool,10.00GB,6.69GB,6.90GB,209.30MB,144,on,80,2
56,yes,no,6.69GB
11,APPVOLA,0,0,DS3K_SASpool,100.00GB,83.17GB,85.18GB,2.01GB,117,on,8
0,256,yes,no,83.17GB
© Copyright IBM Corporation 2011, 2014
Notes:
The CLI refers to thin-provisioned volumes as space-efficient VDisks (SEV). The lssevdiskcopy
command provides a concise view of the capacity utilization data associated with thin-provisioned
volume copies.
One filer value associated with the lssevdiskcopy command is overallocation. Use it periodically to
obtain a list of thin-provisioned volumes whose real capacity consumption is approaching or
exceeding a predetermined threshold.
The example output above provides a list of volumes whose real capacity consumption is 60% or
greater compared to the volume total virtual capacity.
Of course one can wait for the 80% warning threshold alert in the event log or even reset the
volume warning threshold to a lower value.
Uempty
Show dependent volumes if issues with storage
system
1
2
Volumes affected
if storage box is offline
Figure 5-44. Show dependent volumes if issues with storage system SNV13.0
Notes:
Let us explore a feature provided by the GUI in preparation for the SVC Volume Mirroring
discussion.
Click Pools > External Storage then right-click the REDDS3K storage system entry. Select the
Show Dependent Volumes option to view the list of volumes that would be affected if this storage
system were to be taken offline.
This list of volumes, along with their host attachment data, can be used to take preventive action if a
disruptive action needs to be applied to the storage system. Note the APPVOLA volume is on this
list.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-47
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
• Thin Provisioning
• Volume Mirroring
• Real-time Compression
• Comprestimator utility
Notes:
Our next topic discusses volume mirroring.
Uempty
Copy 0 Extent 5
5 GB
Extent 4
Extent 3
Pool1
Extent 2
extent size
Host sees ONE Extent 1
512 MB
5 GB volume Volume Extent 0
(LBA0 – LBAn)
Copy 1 Extent 4
Extent 3
Pool2
Extent 2
Copy has its own:
Extent 1 extent size
• Storage pool
Extent 0 1024 MB
• Virtualization type
• Fully allocated or thin
Notes:
The Volume Mirroring function enables two sets of MDisk extents to be associated with one volume.
These two sets of extents or volume copies can reside in the same or different storage pools (or
storage systems) to provide higher availability to applications.
If a volume is created with two volume copies, these copies use the same virtualization policy
(striped, sequential, image, Thin-Provisioned or fully allocated). A volume copy can also be added
to an existing volume. In this case, the two copies do not have to share the same virtualization
policy. When a volume copy is added, the SVC automatically synchronizes the new copy so that it
contains the same data as the existing copy. The volume is available for host server access during
synchronization process.
The two sets of extents are known as copy 0 and copy 1. Volume copy 0 is always created when a
volume is created. Volume copy 1 is created along with copy 0 if both copies are requested at
volume creation; otherwise copy 1 is added subsequently when an add volume copy request is
processed for an existing volume.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-49
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
Lower Cache
Easy Tier 3
Pool1 Pool2
Virtualization extents extents
Forwarding
Promotes high availability at local site
RAID Expands data migration flexibility
Forwarding Enables non-disruptive conversions
Thin-Provisioned ÅÆ fully allocated
SCSI Initiator
Allows mirrored copy to be split as new
volume
© Copyright IBM Corporation 2011, 2014
Notes:
By default, volume copy 0 is assigned as the primary copy of the volume. Reads and writes always
go through the primary copy. Writes are also sent to volume copy 1 so that synchronization is
maintained between the two volume copies.
If the primary copy is unavailable - for example volume copy 0’s pool became unavailable due to its
storage system has been taken offline - the volume remains accessible to assigned servers. Reads
and writes are handled with volume copy 1. The SVC tracks changed blocks of volume copy 1 and
resynchronize these blocks with volume copy 0 when it becomes available. Reads and writes then
revert back to volume copy 0. It is also possible to set volume copy 1 as the primary copy if desired.
The ability to create a volume copy affords additional management flexibility. A storage controller
might fail or be taken offline for maintenance and not affect application access. A thin-provisioned
volume can easily be converted to fully allocated or vice versa. A volume can be migrated from one
storage pool to another and acquire a different extent size. The original volume copy can then be
deleted. These activities occur within the SVC, totally transparent to attaching servers and user
applications.
Uempty
3 4
Notes:
Return to the example of the Thin-Provisioned APPVOLA volume that had exceeded its space
warning threshold. It has been determined that the volume’s originally allocated capacity is correct
but it should not have been allocated as Thin-Provisioned.
A simple method to convert a thin-provisioned volume copy to fully allocated is to take advantage of
SVC Volume Mirroring to add a fully allocated copy.
To add a mirrored volume copy to an existing volume, right-click the volume entry and select
Volume Copy Actions > Add Mirrored Copy from the pop-up list.
Take another look at the APPVOLA volume entry - note it has an object ID of 11 and resides in the
DS3K_SASpool.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-51
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
Notes:
Three options are in the Volume Type box. This example’s objective is to create a fully allocated
volume copy so the radio button for Generic clicked.
A storage pool that will provide extents for the new volume copy needs to be selected. Target
storage pools with enough free capacity to accommodate the volume copy are displayed by the
GUI. Pool ID 2, the SWV7K_SASpool, is selected for this example.
Study the addvdiskcopy command generated by the GUI - it is adding a volume copy to a volume
whose ID is 11; and extents for the volume copy should be obtained from the pool whose ID is 2.
When the addvdiskcopy command is processed, the APPVOLA volume owns two sets of extents:
The original thin-provisioned volume copy 0 extents are in the DS3K_SASpool, and the newly
added fully allocated volume copy 1 extents are in the SWV7K_SASpool.
The extent size of these two storage pools do have to be identical. SVC Volume Mirroring
processing is independent of the storage pool extent size.
Uempty
Figure 5-50. Volume copy created: Sync the two copies SNV13.0
Notes:
The volume entry is now shown with two copies - copy 0 and copy 1. When a volume is created - it
has, by definition, one set of extents. This set of extents is known as copy 0. The addvdiskcopy
request added copy 1 for this volume.
The asterisk associated with volume copy 0 is used to identify the primary copy - the copy used by
SVC for reads and writes.
The two volume copies need to be synchronized. SVC Volume Mirroring automatically copies the
data of copy 0 to copy 1; while supporting concurrent application reads/writes.
The Running Tasks status bubble indicates that one volume synchronization task is running in the
background. Clicking within the bubble causes the GUI to display its progress.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-53
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
Notes:
The synchronization or background copy rate defaults to 50%, which is set to 2 MBps. It can be
changed by clicking the Edit button in the volume details pane.
Note the capacity bar for copy 1: It is for a fully allocated volume copy.
Volumes are managed by Easy Tier at the volume copy level and is controlled by its associated
pool’s Easy Tier status. Since the Easy Tier status for the SWV7K_SASpool is active, volume copy
1 has automatically became Easy Tier active.
Uempty
Notes:
After clicking the Edit button, the Mirror Sync Rate value becomes an edit box to allow the rate to
be modified. Changing the synchronization rate to 100 increases the background copy rate to 64
MBps.
The chvdisk -syncrate command is generated by the GUI to change the synchronization rate to
100 or 64 MBps.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-55
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
Notes:
The background synchronization rate can be monitored from the Monitoring > Performance view.
The increase in the copy synchronization rate to 100 can be seen in the 64 MBps values for reads
and writes in the MDisks quadrant.
Uempty
copy 1
copy 0
extents
extents
DS3K_SASpool SWV7K_SASpool
Notes:
The Member MDisks tab displays the extent distribution across MDisks for the volume copies in
two different pools. This is the default view. The drop-down list in this view can be used to display
extents associated with each volume copy individually.
The extents in the DS3K_SASpool belong to the thin-provisioned volume copy, copy 0. The fully
allocated volume copy 1 uses extents from the SWV7K_SASpool.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-57
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
1
2
Change copy 1 to
be the primary copy
for volume ID 8
Notes:
The primary copy is used by SVC for both reads and writes. To change volume copy 1 to be the
primary copy, right-click its entry and select Make Primary from the pop-up list.
The GUI generates the chvdisk -primary command to designate volume copy 1 as the primary
copy for the selected volume, volume ID.
A use case for designating volume copy 1 as the primary copy is the migration of a volume to a new
storage system. For a test period, it might be desirable to have both the read and write I/Os directed
at the new storage system of the volume while still maintain a copy in the storage system scheduled
for removal.
Uempty
Notes:
Nothing has changed from a host perspective. I/O operations proceed as normal. SVC Volume
Mirroring is totally transparent to the host.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-59
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
Copy 1 * = reads/writes
I/O #
Notes:
Nothing has changed from a host perspective. I/O operations still proceed as normal.
Uempty
Volumes affected
if storage box is offline
Notes:
Return to Pools > External Storage and right-click the REDDS3K storage system entry. Select the
Show Dependent Volumes option again.
Notice the APPVOLA volume is no longer listed as being dependent on this storage system; even
though its volume copy 0 is using extents from this storage system. Another set of extents in a
different storage system enables the volume to be viable without this storage system.
Mission critical data (volumes) can be identified on a per storage systems basis using this GUI path.
If a volume is dependent, add another copy to the volume in a different pool (different storage
system) before proceeding with maintenance for this storage system.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-61
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
Notes:
Deleting volume copy 0 does not delete the volume; volume copy 1 remains.
Uempty
copy 0
9 11
extents
DS3K_SASpool Copy 1
Copy 0
APPVOL1 APPVOLA
SWV7K_SASpool
Notes:
Observe the command generated by the GUI to delete volume copy 0 is rmvdiskcopy -copy 0 for
volume ID 11. A copy of the volume is removed so the command needs to identify which volume
copy of which volume.
Only volume copy 1 of the volume remains. It is a fully allocated copy: The volume is no longer
Thin-Provisioned.
The deletion of copy 0 has also resulted in a migration of the volume from one storage pool to
another. The extent size of the storage pools do not have to be the same. SVC Volume Mirroring
can be used to migrate volumes between these pools.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-63
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
I/O #
Notes:
Nothing has changed from a host perspective. I/O operations proceed as normal.
Uempty
Copy 0 Copy 1
Fully allocated Thin-Provisioned
Existing data volumes: Go BLUE!
(removes grain-sized strings of zeros)
Notes:
In an analogous manner, fully allocated volumes can be converted to Thin-Provisioned by using
SVC Volume Mirroring to add a Thin-Provisioned copy. After synchronization completes, the fully
allocated volume copy can be deleted.
During the synchronization process, grains containing only zeros are not stored (do not use real
capacity).
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-65
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
Notes:
For this example, APPVOL1 is the party of interest. Based on its host, it has a volume capacity of
100 GB of which 38 GB is free space.
As the data ages and become less active, it might be worthwhile to migrate it to a lower cost
storage ties; and at the same time release the allocated but unused capacity.
Uempty
Convert volume to Thin-Provisioned to release
allocated but unused capacity
Copy 0
APPVOL1
3 4
Figure 5-64. Convert volume to Thin-Provisioned to release allocated but unused capacity SNV13.0
Notes:
A volume copy is to be added to this fully allocated APPVOL1 volume (ID 9) residing in the
SWV7K_SASpool.
© Copyright IBM Corp. 2011, 2014 Unit 5. SVC enhanced features 5-67
Course materials may not be reproduced in whole or in part
without the prior written permission of IBM.
Student Notebook
Notes:
To request a thin-provisioned volume copy, click the radio button for Thin-Provisioned in the
Volume Type box.
The target pool selected is the DS3K_SATApool (pool ID 1). The aim of this exercise is not only to
convert the volume to Thin-Provisioned, but also migrate it to a lower cost storage tier.
Study the addvdiskcopy command generated by the GUI. A thin-provisioned volume copy 1 is
being add to volume ID 9; using extents in pool ID 1 which is the DS3K_SATApool.
Not surprisingly, the GUI uses the Thin-Provision preset attributes for this copy - automatically
expand, real capacity as 2% of virtual capacity, and a warning threshold of 80%. The default grain
size is taken and is not be explicitly coded.