You are on page 1of 74

CMP211-R

Amazon EC2 foundations

Chetan Kapoor
Principal Product Manager, EC2
Amazon Web Services

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon EC2 foundations

Resources Availability Management Purchase Options

Instances Regions and AZs Deployment On Demand


Storage Load Balancing Monitoring Reserved
Networking Auto Scaling Administration Spot
Savings Plan
Amazon Elastic Compute Cloud (Amazon EC2)
Virtual servers in the cloud

EC2 instances Guest 1 Guest 2 Guest n

Hypervisor

Host server

Physical servers in
AWS global regions
Amazon EC2 13+ years ago…

M1

“One size fits all” Pay for what Scale up or down


you use quickly, as needed
270+ instances across
42 instance types
270 +

2017
Journey from then to now

1.7 GHz
1.75 GB
160 GB
250 Mbps
Journey from then to now

1.7 GHz 4.0 GHz


1.75 GB
24 TiB
160 GB
250 Mbps
60 TB

48 TB
d2.8xlarge
100 Gbps
AWS recognized as
a cloud leader for the
9th consecutive year

Gartner, Magic Quadrant for Cloud Infrastructure as a Service, Worldwide, Raj Bala, Bob Gill, Dennis Smith, David Wright, July 2019. ID
G00365830. Gartner does not endorse any vendor, product or service depicted in its research publications, and does not advise technology
users to select only those vendors with the highest ratings. Gartner research publications consist of the opinions of Gartner's research
organization and should not be construed as statements of fact. Gartner disclaims all warranties, expressed or implied, with respect to this
research, including any warranties of merchantability or fitness for a particular purpose. The Gartner logo is a trademark and service mark
of Gartner, Inc., and/or its affiliates, and is used herein with permission. All rights reserved.
Amazon EC2 instance characteristics

Instance *Additional
CPU generation capabilities

Memory
M5d.xlarge
Storage Instance family Instance size

Network performance
Instance type
Broadest choice of processors

Right compute for the right application


Amazon Machine Images (AMIs)

Amazon Marketplace Your machine


maintained maintained images
Broad set of Linux and Managed and AMIs you have created
Windows images maintained by AWS from Amazon EC2
Marketplace partners instances
Kept up to date by
Amazon in each region Can keep private, share
with other accounts, or
Amazon Linux 2 publish to the community
with five years of
long-term support
© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.
General-purpose workloads
Amazon EC2 general-purpose instances

Balance of compute, memory, and network


M5 instances resources. 4:1 memory to vCPU ratio

Baseline level of CPU performance with the


T3 Instances ability to burst above the baseline for workloads
that don’t require sustained performance

Workloads that can scale-out across multiple


A1 Instances cores, fit within memory, run on ARM
instructions
Opportunity: Most instances aren’t very busy
Low utilization

High utilization
Amazon EC2 general-purpose instances

Balance of compute, memory, and network


M5 instances resources. 4:1 memory to vCPU ratio

Baseline level of CPU performance with the


ability to burst above the baseline for
T3 instances workloads that don’t require sustained
performance

Workloads that can scale-out across multiple


A1 Instances cores, fit within memory, run on ARM
instructions
A1 instances powered by AWS Graviton processors

AWS Graviton processor Amazon EC2 A1

Run scale-out and Arm-based


applications in the cloud

AWS Graviton Processor


Up to 45% 64-bit Arm Neoverse cores
cost savings and custom AWS silicon

Flexibility and Lower cost Maximize resource


choice for your efficiency with AWS
workloads Nitro System
Amazon EC2 general-purpose instances

Balance of compute, memory, and network


M5 instances resources. 4:1 memory to vCPU ratio

Baseline level of CPU performance with the


ability to burst above the baseline for
T3 instances workloads that don’t require sustained
performance

Workloads that can scale out across multiple


A1 instances cores, fit within memory, run on ARM
instructions
Announcing AWS Graviton2 processor

Graviton1 processor Graviton2 processor


Announcing Graviton2 based instances

M6g R6g C6g


Memory-intensive workloads
Amazon EC2 memory-optimized instances

Accelerate performance for workloads that


R5 instances process large data sets in memory
8:1 memory to vCPU ratio

For memory-intensive workloads and very


X1 and X1e
large in-memory workloads.
Instances 16:1 and 32:1 memory to vCPU ratio.

Extreme memory needs.


High Memory
Certified to run SAP HANA.
Instances From 6 to 24 TB of memory.
Amazon EC2 memory-optimized instances

Accelerate performance for workloads that


R5 instances process large data sets in memory
8:1 memory to vCPU ratio

For memory-intensive workloads and very


X1 / X1e
large in-memory workloads
instances 16:1 and 32:1 memory to vCPU ratio

Extreme memory needs.


High Memory
Certified to run SAP HANA.
Instances From 6 to 24 TB of memory.
Amazon EC2 memory-optimized instances

Accelerate performance for workloads that


R5 instances process large data sets in memory
8:1 memory to vCPU ratio

For memory-intensive workloads and very


X1 / X1e
large in-memory workloads
instances 16:1 and 32:1 memory to vCPU ratio

High Extreme memory needs


memory Certified to run SAP HANA
instances From 6 to 24 TB of memory
Amazon EC2 instances for SAP HANA
Scale-up options Scale-out options
24
OLTP Scale-Out
(S/4HANA) 48

18
OLAP 12
Scale-Out1 TB
Memory in TB

12
OLAP 5
9 Scale-Out1
0
T
6 B

4 OLAP 100
Scale-Out1 TB
2
.768 1
.244 .384 .488
1 BWoH, BW/4HANA and Datamart
R4 R5 R4 R5 X1 X1 X1e High Memory Instances
Compute-intensive workloads
Batch processing Distributed analytics High-perf computing (HPC)

Ad serving Multiplayer gaming Video encoding


Amazon EC2 compute-optimized instances

High performance at a low price per vCPU ratio


C5 instances 2:1 memory to vCPU ratio

High single thread performance.


z1d Instances Fastest processor in the cloud at 4.0 GHz.
8:1 memory to vCPU ratio.
Amazon EC2 compute-optimized instances

High performance at a low price per vCPU ratio


C5 instances 2:1 memory to vCPU ratio

High single thread performance


z1d instances Fastest processor in the cloud at 4.0 GHz
8:1 memory to vCPU ratio
Storage-intensive workloads
High IO Dense storage
High-perf databases Real-time analytics Big data Data warehousing

Transactional workloads No SQL databases Kafka HDFS

SQL
MapReduce Log processing
Amazon EC2 storage-optimized instances

I3 / I3en I/O optimized for high transaction workloads,


instances low latency workloads

Lowest cost per storage ($/GB).


D2 Instances
Supports high sequential disk throughput.

Designed for applications that require low cost,


high disk throughput and high sequential disk
H1 Instances I/O access to very large data sets. More vCPUs
and memory per TB of disk than D2.
Amazon EC2 storage-optimized instances

I3 / I3en I/O optimized for high transaction workloads,


instances low latency workloads

Lowest cost per storage ($/GB)


D2 D2 instances
Supports high sequential disk throughput

Designed for applications that require low cost,


high disk throughput and high sequential disk
H1 Instances I/O access to very large data sets. More vCPUs
and memory per TB of disk than D2.
Amazon EC2 storage-optimized instances

I3 / I3en I/O optimized for high transaction workloads,


instances low latency workloads

Lowest cost per storage ($/GB)


D2 D2 instances
Supports high sequential disk throughput

Designed for applications that require low cost,


high disk throughput and high sequential disk
H1 instances I/O access to very large data sets
More vCPUs and memory per TB of disk than
D2
Accelerated computing workloads
Applications that benefit from hardware acceleration

Machine learning/AI High-performance Graphics


computing
Image and Video Natural Language Computational Financial and Virtual Graphic 3D Modeling &
Recognition Processing Fluid Dynamics Data Analytics Workstation Rendering

Autonomous Personalization & Computational


Genomics Video Encoding AR/VR
Vehicle Systems Recommendation Chemistry
CPUs vs GPUs vs FPGA vs ASICs for compute acceleration

CPU GPU FPGA ASICs


DRAM DRAM
ALU ALU ALU ALU
Control Control
ALU ALU ALU ALU

Cache Cache
DRAM DRAM
ALU ALU ALU ALU
Control Control
ALU ALU ALU ALU

Cache Cache
DRAM DRAM DRAM DRAM DRAM

• 10s-100s of processing • 1,000s of processing cores • Millions of programmable • Optimized & custom
cores • Pre-defined instruction set digital logic cells design for particular
• Pre-defined instruction set and datapath widths • No predefined instruction use/function
& datapath widths • Highly effective at parallel set or datapath widths • Predefined software
• Optimized for general- execution • Hardware timed execution experience exposed
purpose computing through API
Amazon EC2 accelerated computing instances

GPU compute instance for use cases including deep learning training,
P-Series HPC simulations, financial computing, and batch rendering
P2/P3 instances Feature latest NVIDIA high-end GPUs, including Volta V100

GPU Graphics Instance designed for workloads such as 3D rendering,


remote graphics workstations, video encoding, and AR/VR
G-Series
G3/G4 Instances Feature NVIDIA mid-range GPUs such as Turing T4 GPUs, with GRID
Virtual Workstation features and license

Customer programmable FPGAs that provide dramatic performance


FPGA Instances improvements for applications such as financial computing,
F1 Instances genomics, accelerated search, and image processing
Feature Xilinx Virtex® UltraScale+™ VU9P FPGAs in a single instance.
Programmable via VHDL, Verilog, or OpenCL.
Amazon EC2 accelerated computing instances

GPU compute instance for use cases including deep learning training,
P-Series HPC simulations, financial computing, and batch rendering
P2/P3 instances Feature latest NVIDIA high-end GPUs including Volta V100

GPU graphics instance designed for workloads such as 3D rendering,


remote graphics workstations, video encoding, and AR/VR
G-Series
G3/G4 instances Feature NVIDIA mid-range GPUs such as Turing T4 GPUs, with GRID
Virtual Workstation features and license

Customer programmable FPGAs that provide dramatic performance


FPGA Instances improvements for applications such as financial computing,
F1 Instances genomics, accelerated search, and image processing
Feature Xilinx Virtex® UltraScale+™ VU9P FPGAs in a single instance.
Programmable via VHDL, Verilog, or OpenCL.
Amazon EC2 accelerated computing instances

GPU compute instance for use cases including deep learning training,
P-Series HPC simulations, financial computing, and batch rendering
P2/P3 instances Feature latest NVIDIA high-end GPUs including Volta V100

GPU graphics instance designed for workloads such as 3D rendering,


remote graphics workstations, video encoding, and AR/VR
G-Series
G3/G4 instances Feature NVIDIA mid-range GPUs such as Turing T4 GPUs, with GRID
Virtual Workstation features and license

Customer programmable FPGAs that provide dramatic performance


FPGA instances improvements for applications such as financial computing,
genomics, accelerated search, and image processing
F1 instances
Feature Xilinx Virtex UltraScale+ VU9P FPGAs in a single instance
Programmable via VHDL, Verilog, or OpenCL
Announcing Inf1 instances
Announcing Inf1 instances

40% lower cost-per-inference than any


Amazon EC2 GPU instance

2x higher inference throughput with up to


2,000 TOPS at sub-millisecond latency

High performance and


the lowest cost machine Integration with popular ML frameworks
learning inference in TensorFlow, PyTorch, and MXNet
the cloud
EC2 Bare Metal
Run bare metal workloads on EC2
with all the elasticity, security, scale,
and services of AWS

Designed for workloads that are not virtualized, require specific types of
hypervisors, or have licensing models that restrict virtualization
It all starts with our investments in the Nitro platform
Nitro Card Nitro Security Chip Nitro Hypervisor

Local NVMe storage Integrated into motherboard Lightweight hypervisor


Amazon Elastic Block Storage Protects hardware resources Memory and CPU allocation
Networking, monitoring, and security Bare Metal-like performance
EC2 “instance” host architecture
Server
2012: EC2 “instance” host architecture
Server
Annapurna
2013: EC2 “instance” host architecture
Server
2017: Introducing Nitro architecture
Server
Nitro
2018: Nitro enabling Bare Metal instances
Server
Nitro
Nitro delivers
Broadest and deepest platform choice

Categories Capabilities Options


Choice of processor
General purpose (AWS, Intel, AMD)

Fast processors
Burstable (up to 4.0 GHz)

Compute intensive

Memory intensive
270 +
How do you select the right
High memory footprint
(up to 12 TiB)

Instance storage
Amazon Elastic Block Store

instance to launch and optimize?


(HDD and NVMe)
Storage (High I/O)
Accelerated computing Amazon Elastic Inference for virtually every
Dense storage (GPUs and FPGA) workload and
business need
Networking
GPU compute (up to 100 Gbps)

Graphics intensive Bare Metal


Size
(Nano to 32xlarge)
Announcing

Instance Discovery AWS Compute Optimizer


New search and discovery experience
Machine learning based service that
to easily find EC2 instance types
recommends optimal AWS resources
Quicker and easier for you to find and
Recommends optimal EC2 instances and
compare different instance types
Amazon EC2 Auto Scaling group config
and project costs

Lower costs Optimize performance Get started quickly


Amazon EC2 foundations

Resources Availability Management Purchase Options

Instances Regions and AZs Deployment On Demand


Storage Load Balancing Monitoring Reserved
Networking Auto Scaling Administration Spot
Savings Plan
Amazon EC2 instance store
EC2 instances Local to instance
Non-persistent data store
Data not replicated (by default)
No snapshot support
Instance Store
SSD or HDD

Physical host machine


Amazon EBS
EC2 Block storage as a service
instance
Create, attach, modify through an API
Select storage and compute based on
EBS SSD- your workload
backed
volumes
Detach and attach between instances
gp2 io1
Choice of magnetic and SSD-based
EBS EBS volume types
HDD-
volume backed
volumes Supports snapshots: Point-in-time
st1 sc1
backup of modified volume blocks
EBS
Snapshot
Amazon S3
New EBS performance and security improvements
Encryption by default for EBS Fast Snapshot Restore (FSR) 36% higher EBS-optimized
volumes with opt-in setting bandwidth on C5/C5d, M5/M5d,
R5/R5d instance types

Encrypt all newly created EBS 6x lower recovery time objective Dedicated bandwidth to Amazon EBS
volumes for an account in a region (RTO) 19 Gbps maximum bandwidth, the
Easy to ensure compliance without Skip pre-warming: Instant access to highest across EC2 instances
change to workflows data in snapshot and full
performance upon volume creation
Restore up to 10 volumes
simultaneously
Amazon EC2 foundations

Resources Availability Management Purchase Options

Instances Regions and AZs Deployment On Demand


Storage Load Balancing Monitoring Reserved
Networking Auto Scaling Administration Spot
Savings Plan
Amazon Virtual Private Cloud (Amazon VPC)

Security NAT Flow


groups & ACLs gateway logs

VPC endpoints
Virtual Private Cloud Private and secure connectivity to Amazon S3 and Amazon DynamoDB
Provision a logically isolated
cloud where you can launch
AWS resources into a Amazon S3 Amazon DynamoDB
virtual network

Shared VPC allows multiple accounts to launch their applications into a VPC
Amazon EC2 resources recap

VPC VPC

EBS EBS EBS EBS EBS EBS


Instance
AMI

Virtual machine Running or Amazon EBS


configuration stopped VM Snapshots

Amazon S3
Amazon EC2 foundations

Resources Availability Management Purchase Options

Instances Regions and AZs Deployment On Demand


Storage Load Balancing Monitoring Reserved
Networking Auto Scaling Administration Spot
AWS global platform
SLA of 99.99% availability

AWS Global Infrastructure 191 CloudFront PoPs AWS Global Network


• 22 Regions with 69 Availability Zones • 180 Edge Locations • Redundant 100 GbE network
• 3 Regions coming soon: Indonesia, Italy, and • 11 Regional Edge Caches • 100% encrypted between facilities
South Africa • 73 cities across 33 countries
• Private network capacity between
all AWS Region, except China
Elastic Load Balancing

Load balancer
EC2
instance used to route incoming
requests to multiple
Amazon EC2 instances,
containers, or IP
EC2
ELB instance addresses in your VPC
Elastic Load
Balancing provides
EC2 high-availability
instance by utilizing multiple
Availability Zones
Amazon EC2 Auto Scaling
Dynamically react to changing demand, optimize cost
Fleet management Dynamic scaling
Replace unhealthy instances Scale to demand

CPU
ELB ELB
utilization

EC2 instances EC2 instances

Auto Scaling group Auto Scaling group


Amazon EC2 foundations

Resources Availability Management Purchase Options

Instances Regions and AZs Deployment On Demand


Storage Load Balancing Monitoring Reserved
Networking Auto Scaling Administration Spot
Savings Plan
Launching instances with Launch Templates
Templatize launch requests in order to streamline and simplify future launches

Consistent
Launch experience
parameters
Console
Instance type Simple
EBS volume permissions
AMI ID
Network interface Governance and
Tags CLI best practices
Launch Instances
User data
Block device mapping Increased
Placement
productivity
API
AWS Systems Manager: Operate safely at scale

Cloud
and
Stay patch and configuration compliant
on-premises Automate across accounts and regions
Connect to Amazon EC2 instances via browser and CLI
Track software inventory across accounts
Install agents safely across instances with rate control
Linux
and
Windows
AWS License Manager
Simplified license management for on-premises and cloud

More easily manage licenses from software vendors


Define licensing rules, discover usage, manage access
Gain single view of license across AWS and on-premises
Discover non-compliant software and help prevent misuse
Seamless integration with AWS Systems Manager and
AWS Organizations
Free service for all customers
EC2 foundations

Resources Availability Management Purchase Options

Instances Regions and AZs Deployment On Demand


Storage Load Balancing Monitoring Reserved
Networking Auto Scaling Administration Spot
Savings Plan
Amazon EC2 purchase options

the second a significant more flexibility savings of up to 90%


discount

Savings Plan
Amazon EC2 Reserved Instances pricing

Discount up to 75% off Steady state and 1- and 3-year terms


the On-Demand price committed usage

Payment flexibility with Convertible RIs Reserve capacity or opt for


3 upfront payment options Change instance family, flexibility across AZs and
(all, partial, none) OS, tenancy, and payment instance sizes

On-Demand capacity reservations: Manage capacity and RI decisions independently


1-Year Convertible RIs
Simplifying purchasing with Savings Plans
Flexible purchase option that offers savings of up to 72% on Amazon EC2 and AWS
Fargate usage

Easy Significant Flexible


to use savings

Same great prices as EC2 RIs with more flexibility


Amazon EC2 Spot pricing
Spare Amazon EC2 capacity at savings of up to 90% over On-Demand

Lean on Spot for these workloads!

Faster results
Increase throughput up to 10x
while staying in budget

Big data CI/CD Web services HPC

Spot is ideal for:


Easy to use Or containerized workloads ☑Fault-tolerant
Launch through AWS services ☑Flexible
(ex. Amazon ECS, Amazon EKS,
AWS Batch, Amazon EMR) or ☑Loosely coupled
integrated third-parties ☑Stateless workloads
To optimize Amazon EC2, combine purchase options

Spot for fault-tolerant,


flexible, stateless workloads

On-Demand

RIs or Savings Plan


Amazon EC2 foundations

Resources Availability Management Purchase Options

Instances Regions and AZs Deployment On Demand


Storage Load Balancing Monitoring Reserved
Networking Auto Scaling Administration Spot
Thank you!

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.
© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.

You might also like