The AWS Compute platform has expanded EC2 instance types including FPGA and new GPU instances. There are also other ways to run workloads in AWS including Lambda (serverless), ECS (managed Docker), and AWS Batch (batch computing). This session will cover the newest instance types in EC2 and review AWS Lambda, ECS, and Batch. Learn More: https://aws.amazon.com/government-education/

1. © 2017, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Mark Smith, AWS Solutions Architect Manager
June 14, 2017
AWS Compute Overview
Servers, Containers, Serverless, and Batch

2. What to expect from this session?
• New developments in the Amazon EC2 platform
• Compute
• Storage
• Networking
• Review of additional Compute options in AWS
• AWS Lambda
• Amazon ECS
• AWS Batch

3. AWS Pace of Innovation
2011 2012 2013 2014 2015 2016
80+
160
280
516
722
1017

5. Broad Set of Compute Instance Types
M4
General
purpose
Compute
optimized
C4
C3
Storage and I/O
optimized
I3
G2
GPU or FPGA
enabled
Memory
optimized
X1
P2
F1
R4
R3
C5
I2
D2

6. 2006 2008 2010 2012 2014 2016
m1.small
m1.large
m1.xlarge
c1.medium
c1.xlarge
m2.xlarge
m2.4xlarge
m2.2xlarge
cc1.4xlarge
t1.micro
cg1.4xlarge
cc2.8xlarge
m1.medium
hi1.4xlarge
m3.xlarge
m3.2xlarge
hs1.8xlarge
cr1.8xlarge
c3.large
c3.xlarge
c3.2xlarge
c3.4xlarge
c3.8xlarge
g2.2xlarge
i2.xlarge
i2.2xlarge
i2.4xlarge
i2.4xlarge
m3.medium
m3.large
r3.large
r3.xlarge
r3.2xlarge
r3.4xlarge
r3.8xlarge
t2.micro
t2.small
t2.med
c4.large
c4.xlarge
c4.2xlarge
c4.4xlarge
c4.8xlarge
d2.xlarge
d2.2xlarge
d2.4xlarge
d2.8xlarge
g2.8xlarge
t2.large
m4.large
m4.xlarge
m4.2xlarge
m4.4xlarge
Amazon EC2 Instances History
t2.nano
2018
x1.16xlarge
x1.32xlarge
f1.2xlarge
f1.16xlarge
p2.xlarge
p2.8xlarge
p2.16xlarge
t2.xlarge
t2.2xlarge
m4.10xlarge
m4.16xlarge
i3.large
i3.xlarge
i3.2xlarge
i3.4xlarge
i3.8xlarge
i3.16xlarge
r4.large
r4.xlarge
r4.2xlarge
r4.4xlarge
r4.8xlarge
r4.16xlarge

7. Elastic Network Adapter (ENA)
• New PCI network device developed for EC2
• Supported on F1, I3, R4, C1, X1, P2, and m4.16xlarge
instance types
• Ability to scale across variety of bandwidths
• 10 and 20 Gbps instance types today
• Device interface supports up to 400 Gbps networking

8. TCP Bandwidth
0
5000
10000
15000
20000
25000
64 128 256 512 1024 2048 4096 8192
Mbps
TCP Packet Size
cc2.8xlarge c3.8xlarge
m4 m4.16xlarge m4.16xlarge Multistream

9. Instance Type Dedicated EBS Throughput
(Mbps)*
Max 16K
IOPS**
m1.xlarge 1000 8000
m2.4xlarge 1000 8000
m3.2xlarge 1000 8000
g2.2xlarge 1000 8000
c3.4xlarge 2000 16000
i2.4xlarge 2000 16000
Instance type Max. bandwidth
(Mbps)*
Max. IOPS (16 KB
I/O size)**
c4.8xlarge 4,000 32,000
d2.8xlarge 4,000 32,000
m4.16xlarge 10,000 65,000
p2.16xlarge 10,000 65,000
x1.32xlarge 10,000 65,000
r4.16xlarge 12,000 65,000
i3.16xlarge 14,000 65,000
f1.16xlarge 14,000 75,000
2014 2017
4x Increase IOPS
7x Increase throughput
Dedicated EBS
Performance

10. R3
244 GiB
memory
DDR3
32 vCPUs
L3 cache
R3: Memory-Intensive
Applications on EC2
High performance databases
Distributed memory caches
In-memory analytics
Genome assembly & analysis

11. R4: Next Generation
Memory-Optimized
Instances
488 GiB
memory
DDR4
64 vCPUs
L3 cache
R4
Generally Available Today
High performance databases
Distributed memory caches
In-memory analytics
Genome analysis
2X memory
2X fast
1.8X size
2X vCPUs
R3
244 GiB
memory
DDR3
32 vCPUs
L3 cache

12. I2
365k IOPS
244 GiB
memory
32 vCPUs
6.4 TB
SSD
I2: Running I/O-Intensive
Applications on EC2
NoSQL databases
(Cassandra and MongoDB)
Scale out transactional
databases
Data warehousing
Hadoop
Cluster file systems

13. I3: Next Generation
I/O-Intensive Instances
3.3 million
IOPS
488 GiB
memory
64 vCPUs
15.2 TB
NVMe SSD
I3
Coming Soon
NoSQL databases (Cassandra
and MongoDB)
Scale out transactional
databases
Data warehousing
Hadoop
Cluster file systems
I2
365k IOPS
32 vCPUs
244 GiB
memory
6.4 TB
SSD
9X IOPS
2X memory
2.3X storage
2X vCPUs

14. C4
36 vCPUs
“Haswell”
60 GiB
memory
4 Gbps
to EBS
C4: Compute-Optimized
Instances
Web servers
Batch processing
Distributed analytics
High performance science
& engineering
Ad serving

15. C5: Next Generation
Compute-Optimized
Instances with
Intel “Skylake”
Coming Soon
C4
36 vCPUs
“Haswell”
60 GiB
memory
4 Gbps
to EBS
AVX 512
72 vCPUs
“Skylake”
144 GiB memory
C5
12 Gbps to EBS
2X vCPUs
2X performance
3X throughput
2.4X memory

16. X1
128 vCPUs
20 Gbps
Network
2 TB
memory
10 Gbps
to EBS
X1: Memory-Intensive
Instances
In-memory databases
Big data processing
HPC workloads
Machine learning

17. 128 vCPUs
20 Gbps
network
4 TB memory
14 Gbps
X1e
1.4X throughput
2X memory
X1
128 vCPUs
20 Gbps
network
2 TB memory
10 Gbps
To EBS
X1e: Memory-Intensive
Instances
Coming Soon
8TB and 16TB
Planned for 2017 and 2018

18. NVIDIA Tesla
GPU Card
Xilinx
UltraScale+
FPGA
F1: FPGA-accelerated computing
GPU and FPGA for Accelerated Computing
P2: GPU-accelerated computing
70 teraflops 30X Acceleration

19. Control
ALU
ALU
Cache
DRAM
ALU
ALU
CPU
(one core)
FPGA
DRAM DRAM
GPU
Each FPGA in
F1 has more
than 2M of
these cells
Each GPU
in P2 has
2880 of
these
cores
DRAM
Parallel Processing in GPUs and FPGAs
BlockRAM
BlockRAM
DRAM DRAM

20. FPGAs at Work
Ryft Cloud for F1 accelerated Elasticsearch
Edico Genome’s DRAGEN Bio-IT for real-time gene analysis
Falcon Computing Merlin C/C++ to FPGA compiler

21. C o m p u t e

22. Virtual Machines Containers Serverless
A S p e c t r u m O f C o m p u t e

23. Packaging Updates Execution Run time Unit of Cost
VMs AMI Patching
Multi-threaded, multi-
task
Hours to months Per VM per hour
Containers Container File Versioning
Multi-threaded,
single-task
Minutes to days Per VM per hour
Lambda Code Versioning
Single-threaded,
single-task
Microseconds to
seconds
Per memory/second
Per request
A S p e c t r u m O f C o m p u t e

24. Virtual Machines
A S p e c t r u m O f C o m p u t e

25. EBSAuto ScalingVPC RDSELB
S t r o n g V M E c o s y s t e m

26. Containers
A S p e c t r u m O f C o m p u t e

27. What are containers?
OS virtualization
Process isolation
Images
AutomationServer
Guest OS
Bins/Libs Bins/Libs
App2App1

28. Container advantages
Portable
Flexible
Fast
EfficientServer
Guest OS
Bins/Libs Bins/Libs
App2App1

29. Going from this…
Server
Guest OS
Bins/Libs Bins/Libs
App2App1

30. Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
Server
Guest OS
to this… is hard!

31. High
availability
Infrastructure
management
Security Task
scheduling
Pipeline
integration
Container
management
Service
discovery
Resource
access
T h e C h a l l e n g e s O f C o n t a i n e r B a s e d O p e r a t i o n s

32. Amazon ECS
Managed platform for:
C l u s t e r M a n a g e m e n t C o n t a i n e r O r c h e s t r a t i o n D e e p A W S I n t e g r a t i o n

33. D e e p I n t e g r a t i o n W i t h A W S P l a t f o r m
ELBContainer
Auto Scaling
VPCIAMInstance
Auto Scaling
CloudTrailCloudWatch
Metrics
CloudWatch Logs CloudWatch
Events
X-Ray

34. ECS Common Use Cases
Microservices
Batch Processing
PaaS
CICD

35. You can now determine scheduling policies based on:
A M I I D Availability
Zone
Instance
type
Distinct
instances
C u s t o m
T a s k P l a c e m e n t E n g i n e

36. B i n p a c k i n g S p r e a d A f f i n i t y
D i s t i n c t
I n s t a n c e
T a s k P l a c e m e n t S t r a t e g i e s

37. Real-time notifications of current state
C o n t a i n e r
i n s t a n c e
e v e n t s
Task
events
E v e n t S t r e a m

38. Companies
using ECS

39. 21 Services
1.3 Billion
requests per day

40. C4 R3 M4R3 R3
R3 R3 R3
M4 M4
M4 M4 M4
C4 C4
C4 C4 C4
Map Service Search Service Directions Service

41. C4
ECS Cluster
R3 R3 R3
R3 R3 R3
M4
M4
M4 M4
M4 M4
C4 C4
C4 C4 C4
Map Service Search ServiceDirections Service
Spot Fleet
C4
C4
R3
R3Stability On Spot with
Instance Diversity

42. Increased
Efficiency with
Instance Packing
Map
service
CPU 55%
Mem 5%
Search
service
CPU 25%
Mem 75%
Combined
services
with ECS
CPU 80%
Mem 80%
25% fewer instances

43. How ECS
Enables
AWS Batch Architecture
ECS Cluster
AWS Batch Compute Environment
EC2 EC2 EC2 EC2
EC2 EC2 EC2 EC2
Job
Queue
EC2 Auto Scaling or Spot Fleet
ECS
RunTask
ECS
RunTask
AWS Batch
Scheduler
Jobs
AWS Customer Account

44. Serverless
AWS Lambda
A S p e c t r u m O f C o m p u t e

45. N o S e r v e r i s E a s i e r t o M a n a g e
T h a n N o S e r v e r

46. S t a t e O f A W S L a m b d a
Languages Integrations
Node.js (JavaScript)
Java (Java 8 compatible)
Python
A m a z o n A P I G a t e w a y
AWS CodeCommit
AWS CloudFormation
Amazon Simple Notification Service
Amazon Simple Email Service
Amazon CloudWatch
Amazon Kinesis Streams
Amazon DynamoDB
Amazon RDS
Amazon Aurora
A m a z o n C o g n i t o
A m a z o n S 3
AWS Config

47. I A M R o l e s V P C S u p p o r t
S e r v e r l e s s S e c u r i t y

48. AWS X-Ray
Amazon API
Gateway
Amazon
DynamoDB
Amazon
CloudSearch
AWS Lambda
Amazon S3
Amazon SQS
Amazon SNS
Amazon Athena Amazon SES
Amazon
Elasticsearch
Service
AWS CloudFrontAmazon
ElastiCache
Amazon
Kinesis Streams
T h e r e A r e N o C a t t l e , T h e r e I s O n l y T h e H e r d

49. T h o m s o n
R e u t e r s
p r o c e s s e s
4 , 0 0 0 r e q u e s t s
p e r s e c o n d
F I N R A
p r o c e s s e s h a l f
a t r i l l i o n
v a l i d a t i o n s o f
s t o c k t r a d e s
d a i l y
H e a r s t r e d u c e d t h e
t i m e t o i n g e s t a n d
p r o c e s s d a t a f o r i t s
a n a l y t i c s p i p e l i n e
b y 9 7 %
V e v o c a n
h a n d l e s p i k e s
o f 8 0 x n o r m a l
t r a f f i c
E x p e d i a t r i g g e r s
1 . 2 b i l l i o n
L a m b d a r e q u e s t s
e a c h m o n t h
M a s s i v e S c a l e A c h i e v e d

50. Connective Tissue Of AWS Environments
L a m b d a A s E x t e n s i b i l i t y M e c h a n i s m
CloudWatch
Events
ELB LifeCycle
Hooks
Config Rules C o d e D e p l o y
E v e n t s
C o n s o l e
S i g n - i n s
S c h e d u l e d
J o b s
Cognito Your
User Pools
T r u s t e d
A d v i s o r
C l o u d F o r m a t i o nA u t o S c a l i n g
E v e n t s
E c o s y s t e m

51. Connective Tissue Of AWS Environments
A m a z o n L e x A W S G r e e n g r a s s A W S S n o w b a l l
E d g e
A l e x a S k i l l s K i t
L a m b d a A s E x t e n s i b i l i t y M e c h a n i s m

52. Virtual Machines Containers Serverless
F i r s t - C l a s s P r i m i t i v e s W i t h D e e p A W S I n t e g r a t i o n

53. Thank You!