Cloud computing gives you a number of advantages, such as being able to scale your application on demand. As a new business looking to use the cloud, you inevitably ask yourself, "Where do I start?" Join us in this session to understand best practices for scaling your resources from zero to millions of users. We will show you how to best combine different AWS services, make smarter decisions for architecting your application, and best practices for scaling your infrastructure in the cloud.

1. © 2014 Amazon.com, Inc. and its affiliates. All rights reserved. May not be copied, modified, or distributed in whole or in part without the express consent of Amazon.com, Inc.© 2014 Amazon.com, Inc. and its affiliates. All rights reserved. May not be copied, modified, or distributed in whole or in part without the express consent of Amazon.com, Inc.
Scaling on AWS for the
First 10 Million Users
Mohan Vedula, Enterprise Solutions Architecture
Brett Francis, Enterprise Solutions Architecture
March 26, 2014

2. Scaling on AWS for the First 10 Million Users
• US:
– Brett Francis – brettf@amazon.com
– Mohan Vedula – mohanv@amazon.com
• YOU: Here to learn more about scaling
infrastructure on AWS
• TODAY: about best practices and things to think
about when building for large scale

3. Scaling to 10M Users: A Story in Four Parts
• Intro and Initial Steps
• Building Blocks
• Tools and Monitoring
• 10M Users and Beyond

4. So how do we scale?

7. a lot of things to read

8. not where we want to start
a lot of things to read

9. us today

10. Auto Scaling is a tool and a
destination. It’s not the single
thing that fixes everything.

11. What do we
need first?

12. Some basics…

13. Regions
US-WEST (Oregon)
EU-WEST (Ireland)
ASIA PAC (Tokyo)
US-WEST (N. California)
SOUTH AMERICA (Sao Paulo)
US-EAST (Virginia)
AWS GovCloud (US)
ASIA PAC
(Sydney)
ASIA PAC
(Singapore)
CHINA (Beijing)

14. Availability Zones
US-WEST (Oregon)
EU-WEST (Ireland)
ASIA PAC (Tokyo)
US-WEST (N. California)
SOUTH AMERICA (Sao Paulo)
US-EAST (Virginia)
AWS GovCloud (US)
ASIA PAC
(Sydney)
ASIA PAC
(Singapore)
CHINA (Beijing)

15. Edge Locations

16. • $7B+ retail business
• 8,000+ employees
• A whole lot of servers
Every day, AWS adds enough
server capacity to power that
entire $7B enterprise
2004 2014

17. Compute
Storage &
Content
Delivery
AWS Global Infrastructure
Database
App Services
Deployment & Administration
Networking

18. Compute
Storage &
Content
Delivery
AWS Global Infrastructure
Database
App Services
Deployment & Administration
Networking
Amazon
CloudSearch
Amazon
SQS
Amazon
SNS
Amazon
Elastic
Transcoder
Amazon SWF
Amazon
SES
Amazon
DynamoDB
Amazon
RDS
Amazon
ElastiCache
Amazon
RedShift
AWS Storage
Gateway
Amazon S3
Amazon
Glacier
Amazon
CloudFrontAmazon
EC2
Amazon
EMR
Amazon
VPC
Amazon
Route 53
AWS
Direct
Connect
Amazon
Kinesis
Amazon
CloudWatch AWS IAM AWS
CloudFormation
Amazon Elastic
Beanstalk
AWS
Data
Pipeline
AWS
OpsWorks
AWS
CloudTrail

19. So let’s start from day
one, user one ( you )

20. Day One, User One:
• A single EC2 instance
– With a full stack on this host
• Web app
• Database
• Management
• etc.
• A single Elastic IP address
• Amazon Route 53 for DNS
EC2
instance
Elastic IP
address
Amazon
Route 53
User

21. “We’re gonna need a bigger box”
• Simplest approach
• Can now leverage PIOPs
• High I/O instances
• High memory instances
• High CPU instances
• High storage instances
• Easy to change instance sizes
• Will hit an endpoint eventually
m3.xlarge
m1.small
i2.4xlarge

22. “We’re gonna need a bigger box”
m3.xlarge
m1.small
i2.4xlarge
• Simplest approach
• Can now leverage PIOPs
• High I/O instances
• High memory instances
• High CPU instances
• High storage instances
• Easy to change instance sizes
• Will hit an endpoint eventually

23. Day One, User One:
• We could potentially get
to a few hundred to a few
thousand depending on
application complexity
and traffic
• No failover
• No redundancy
• Too many eggs in one
basket
EC2
instance
Elastic IP
address
Amazon
Route 53
User

24. Day One, User One:
• We could potentially get
to a few hundred to a few
thousand depending on
application complexity
and traffic
• No failover
• No redundancy
• Too many eggs in one
basket
EC2
instance
Elastic IP
address
Amazon
Route 53
User

25. Day Two, User >1:
First, let’s separate out
our single host into
more than one:
• Web
• Database
– Make use of a database
service?
Web
instance
Database
instance
Elastic IP
address
Amazon
Route 53
User

26. Self-Managed Fully-Managed
Database server
on Amazon EC2
Your choice of
database running on
Amazon EC2
Bring Your Own
License (BYOL)
Amazon
DynamoDB
Managed NoSQL
database service
using SSD storage
Seamless scalability
Zero administration
Amazon RDS
Microsoft SQL,
Oracle, MySQL or
PostgreSQL as a
managed service
Flexible licensing
BYOL or License
Included
Amazon
Redshift
Massively parallel,
petabyte-scale, data
warehouse service
Fast, powerful and
easy to scale
Database Options

27. Scaling to 10M Users: A Story in Four Parts
• Intro and Initial Steps
• Building Blocks
• Tools and Monitoring
• 10M Users and Beyond

28. But how do I choose
the DB technology I
need? SQL? NoSQL?

29. Some folks won’t like
this. But…

30. Start with SQL
databases

31. Unless you already
have NoSQL skilled
staff…

32. Start with SQL
databases

33. Why start with SQL?
• Established and well-worn technology
• Lots of existing code, communities, books, background,
tools, etc.
• You aren’t going to break SQL DBs in your first 10 million
users. No really, you won’t*.
• Clear patterns to scalability
* Unless you are manipulating data at MASSIVE scale; even then, SQL will have a
place in your stack

34. AH HA! You
said “massive
amounts”, I
will have
massive
amounts!

35. If your usage is such that you will be
generating several TB ( >5 ) of data
in the first year OR have an
incredibly data-intensive workload…
you might need NoSQL

36. Regardless, why NoSQL?
• Super low latency applications
• Metadata driven datasets
• Highly non-relational data
• Need schema-less data constructs*
• Massive amounts of data (again, in the TB range)
• Rapid ingest of data ( thousands of records/sec )
• Already have skilled staff
*Need != “it is easier to do dev without schemas”

37. When NoSQL = Yes…
investigate use of DynamoDB

38. Amazon Dynamo DB
• Managed, provisioned throughput
NoSQL database
• Fast, predictable performance
• Fully distributed, fault tolerant
architecture
• Considerations for non-uniform
data
Feature Details
Provisioned
throughput
Dial up or down provisioned
read/write capacity
Predictable
performance
Average single digit millisecond
latencies from SSD-backed
infrastructure
Strong
consistency
Be sure you are reading the
most up-to-date values
Fault
tolerant
Data replicated across
Availability Zones
Monitoring Integrated to Amazon
CloudWatch
Secure Integrates with AWS Identity
and Access Management
(AWS IAM)
Amazon
EMR
Integrates with Amazon EMR
for complex analytics on large
datasets

39. But back to the main
path… Let’s see how
far SQL at the core
can grow

40. User >100:
First, let’s separate out
our single host into
more than one:
• Web
• Database
– Use Amazon RDS to make
your life easier
Web
instance
Elastic IP
address
RDS DB
instance
Amazon
Route 53
User

41. User > 1000:
Next, let’s address the
lack of failover and
redundancy issues:
• Elastic Load Balancing
• Another web instance
– In another Availability Zone
• Enable Amazon RDS Multi-AZ
Web
instance
Amazon RDS DB instance
Active (Multi-AZ)
Availability Zone Availability Zone
Web
instance
Amazon RDS DB instance
Standby (Multi-AZ)
Elastic Load
Balancing
Amazon
Route 53
User

42. • Create highly-scalable applications
Feature Details
Available Load balance across instances in multiple
Availability Zones
Health checks Automatically checks health of instances and
takes them in or out of service
Session stickiness Route requests to the same instance
Secure sockets layer Supports SSL offload from web and application
servers with flexible cipher support
Monitoring Publishes metrics to Amazon CloudWatch
Elastic Load
Balancing
Elastic Load Balancing

43. Scaling this horizontally
and vertically
will get us pretty far
( 10s-100s of thousands )

44. User >10ks-100ks:
RDS DB Instance
Active (Multi-AZ)
Availability Zone Availability Zone
RDS DB Instance
Standby (Multi-AZ)
Elastic Load
Balancing
RDS DB Instance
Read Replica
RDS DB Instance
Read Replica
RDS DB Instance
Read Replica
RDS DB Instance
Read Replica
Web
instance
Web
instance
Web
instance
Web
instance
Web
instance
Web
instance
Web
instance
Web
instance
Amazon
Route 53
User

45. This will take us pretty far,
honestly, but we care
about performance and
efficiency, so let’s clean up
with some components and
services

46. Shift some load around:
Think components and
services:
• Move static content from the
web instance to Amazon S3
and Amazon CloudFront
• Move session/state and DB
caching to Amazon
ElastiCache or Amazon
DynamoDB
• More on services later…
Web
instance
RDS DB Instance
Active (Multi-AZ)
Availability Zone
Elastic Load
Balancing
Amazon S3
Amazon
CloudFront
Amazon
Route 53
User
ElastiCache
Amazon
DynamoDB

47. Working with Amazon S3
• Object-based storage for the web
• Designed for 11 9s of durability
• Good for things like:
– Static assets (css, js, images, videos)
– Backups
– Logs
– Ingest of files for processing
• “Infinitely scalable”
• Supports fine-grained permission control
• Ties in well with CloudFront
• Ties in with Amazon EMR
• Acts as a logging endpoint for Amazon
S3/CloudFront/Billing
• Supports encryption at transit and at rest
• Reduced redundancy 1/3 cheaper
• Amazon Glacier for super long term
storage

48. CloudFront
Amazon CloudFront is a web service for
scalable content delivery:
• Cache static content at the edge for faster delivery
• Helps lower load on origin infrastructure
• Dynamic and static content
• Streaming video
• Zone apex support
• Custom SSL certificates
• Low TTLs (as short as 0 seconds)
• Lower costs for origin fetches (between Amazon
S3 / Amazon EC2 and CloudFront)
• Optimized to work with Amazon EC2, Amazon S3,
Elastic Load Balancing, and Amazon Route 53
ResponseTime
ServerLoad
ResponseTime
Server
Load
ResponseTime
Server
Load
No CDN CDN for static
content
CDN for static &
dynamic content
0
10
20
30
40
50
60
70
80
8:00
AM
9:00
AM
10:00
AM
11:00
AM
12:00
PM
1:00
PM
2:00
PM
3:00
PM
4:00
PM
5:00
PM
6:00
PM
7:00
PM
8:00
PM
9:00
PM
VolumeofData
Delivered(Gbps)

49. Shift some load around:
Let’s lighten the load on our
web and database
instances:
• Move static content from the
web instance to Amazon S3
and CloudFront
• Move session/state and DB
caching to ElastiCache or
Amazon DynamoDB
Web
instance
RDS DB Instance
Active (Multi-AZ)
Availability Zone
Elastic Load
Balancer
Amazon S3
Amazon
CloudFront
Amazon
Route 53
User
ElastiCache
Amazon
DynamoDB

50. Shift some load around:
Let’s lighten the load on our
web and database
instances:
• Move static content from the
web instance to Amazon S3
and CloudFront
• Move dynamic content from
the load balancer to
CloudFront
• Move session/state and DB
caching to ElastiCache or
Amazon DynamoDB
Web
instance
RDS DB Instance
Active (Multi-AZ)
Availability Zone
Elastic Load
Balancer
Amazon S3
Amazon
CloudFront
Amazon
Route 53
User
ElastiCache
Amazon
DynamoDB

51. Shift some load around:
Let’s lighten the load on our
web and database
instances:
• Move static content from the
web instance to Amazon S3
and CloudFront
• Move dynamic content from
the load balancer to
CloudFront
• Move session/state and DB
caching to ElastiCache or
Amazon DynamoDB
Web
instance
RDS DB Instance
Active (Multi-AZ)
Availability Zone
Elastic Load
Balancer
Amazon S3
Amazon
CloudFront
Amazon
Route 53
User
ElastiCache
Amazon
DynamoDB

52. Scaling to 10M Users: A Story in Four Parts
• Intro and Initial Steps
• Building Blocks
• Tools and Monitoring
• 10M Users and Beyond

53. Now that our web tier is
much more lightweight,
we can revisit the
beginning of our talk…

54. Auto Scaling!

55. Automatic resizing of compute clusters
based on demand
Trigger auto-scaling policy
Feature Details
Control Define minimum and maximum instance pool
sizes and when scaling and cool down occurs
Integrated to Amazon
CloudWatch
Use metrics gathered by CloudWatch to drive
scaling
Instance types Run Auto Scaling for On-Demand and Spot
Instances; compatible with VPC
aws autoscaling create-auto-scaling-
group
--auto-scaling-group-name MyGroup
--launch-configuration-name MyConfig
--min-size 4
--max-size 200
--availability-zones us-west-2c
Auto Scaling
Amazon
CloudWatch

56. Sunday Monday Tuesday Wednesday Thursday Friday Saturday
Typical weekly traffic to Amazon.com

57. Sunday Monday Tuesday Wednesday Thursday Friday Saturday
Typical weekly traffic to Amazon.com
Provisioned capacity

58. November traffic to Amazon.com
November

59. November traffic to Amazon.com
Provisioned capacity
November

60. November traffic to Amazon.com
76%
24%
Provisioned capacity
November

61. November traffic to Amazon.com
November

62. Auto Scaling
lets you do this!

63. Auto Scaling can help from
one instance to thousands
and back down

64. User >500k+:
Availability Zone
Amazon
Route 53
User
Amazon S3
Amazon
CloudFront
Availability Zone
Elastic Load
Balancing
Amazon
DynamoDBRDS DB Instance
Read Replica
Web
instance
Web
instance
Web
instance
ElastiCache RDS DB Instance
Read Replica
Web
instance
Web
instance
Web
instance
ElastiCacheRDS DB Instance
Standby (Multi-AZ)
RDS DB Instance
Active (Multi-AZ)

66. Use Tools:
Managing your infrastructure will become an ever
increasing important part of your time. Use tools to
automate repetitive tasks.
• Tools to manage AWS resources
• Tools to manage software and configuration on your
instances
• Automated data analysis of logs and user actions

67. AWS Application Management Solutions
AWS
Elastic Beanstalk
AWS
OpsWorks
AWS
CloudFormation
Amazon EC2
Convenience Control
Higher-level services Do it yourself

68. Host-Based Configuration Management
Two big players:
– Opscode Chef
– PuppetLabs Puppet
• Both do more or less the same thing
• Both have syntax that isn’t too dissimilar
• Use HBCM with one of the tools from the previous slide
• Spend the time required to learn them
• Can’t scale easily without HBCM
• Growing in popularity: Salt, Ansible

69. User >500k+:
You’ll potentially start to run into issues with speed and
performance of your applications:
• Have monitoring/metrics/logging in place
– If you can’t build it internally, outsource it! (3rd party SaaS)
• Pay attention to what customers are saying works well
• Squeeze as much performance as you can out of each
service/component

70. HOST
LEVEL
METRICS
AGGREGATE
LEVEL
METRICS
LOG
ANALYSIS
EXTERNAL
SITE
PERFORMANCE

71. Not having proper
monitoring/metrics is like
flying a plane with an eye
mask on in a thunderstorm.
Oh, and your wing is on fire.

73. AWS Marketplace & Partners Can Help
• Customer can find, research,
and buy software
• Simple pricing, aligns with
Amazon EC2 usage model
• Launch in minutes
• AWS Marketplace billing
integrated into your AWS
account
• 1300+ products across 20+
categories
Learn more at: aws.amazon.com/marketplace

74. Scaling to 10M Users: A Story in Four Parts
• Intro and Initial Steps
• Building Blocks
• Tools and Monitoring
• 10M Users and Beyond

75. There are further
improvements to be
made in breaking apart
our web/app layer

77. SOA = Service Oriented Architecture

78. SOA’ing
Move services into their own
tiers/modules. Treat each of these
as 100% separate pieces of your
infrastructure and scale them
independently.
Amazon.com and AWS do this
extensively! It offers flexibility and
greater understanding of each
component.

79. Loose coupling sets you free!
• The looser they're coupled, the bigger they scale
– Independent components
– Design everything as a black box
– Decouple interactions
– Favor services with built-in redundancy and scalability rather than
building your own
Controller A Controller B
Controller A Controller B
Q Q
Tight coupling
Use Amazon SQS for buffers
Loose coupling

80. Loose coupling + SOA = winning
Examples:
• Email
• Queuing
• Transcoding
• Search
• Databases
• Monitoring
• Metrics
• Logging
Amazon
CloudSearch
Amazon SQSAmazon SNS
Amazon Elastic
Transcoder
Amazon SWF
Amazon SES
In the early days, if someone has a service for it already,
opt to use that instead of building it yourself.
DON’T RE-INVENT THE WHEEL

81. On re-inventing the wheel…
If you find yourself writing
your own: queue, DNS server,
database, storage system,
monitoring tool

82. Take a deep breath and
stop it. Now.

83. Back to SOA

84. User >1mil+:
Reaching a million and above is going to require some bit
of all the previous things:
• Multi-AZ
• Elastic Load Balancing between tiers
• Auto Scaling
• Service-oriented architecture
• Serving content smartly (Amazon S3/CloudFront)
• Caching off DB
• Moving state off tiers that auto-scale

85. User >1mil+:
RDS DB Instance
Active (Multi-AZ)
Availability Zone
Elastic Load
Balancing
RDS DB Instance
Read Replica
RDS DB Instance
Read Replica
Web
instance
Web
instance
Web
instance
Web
instance
Amazon
Route 53
User
Amazon S3
Amazon
CloudFront
Amazon
DynamoDB
Amazon SQS
ElastiCache
Worker
instance
Worker
instance
Amazon
CloudWatch
Internal app
instance
Internal app
instance
Amazon SES

86. The next big steps

87. User >5mil – 10mil:
You’ll potentially start to run into issues with your database
around contention on the write master.
How can you solve it?
• Federation ~ splitting into multiple DBs based on function
• Sharding ~ splitting one data set up across multiple hosts
• Moving some functionality to other types of DBs (NoSQL)

88. Shifting functionality to NoSQL
• Similar in a sense to federation
• Again, review the earlier points to determine need of
NoSQL vs SQL
• Leverage hosted services like Amazon DynamoDB
• Some use cases:
– Leaderboards/scoring
– Rapid ingest of clickstream/log data
– Temporary data needs ( cart data )
– “Hot” tables
– Metadata/lookup tables
Amazon
DynamoDB

89. …and there you have it.
10 Million

90. A Quick Review

91. Review
• Multi-AZ your infrastructure
• Make use of self-scaling services
– Elastic Load Balancing, Amazon S3, Amazon SNS, Amazon SQS,
Amazon SWF, Amazon SES, etc.
• Build in redundancy at every level
• Most likely start with SQL
• Cache data both inside and outside your
infrastructure
• Use automation tools in your infrastructure

92. Review (cont)
• Make sure you have good
metrics/monitoring/logging tools in place
• Split tiers into individual services (SOA)
• Use Auto Scaling when you’re ready for it
• Don’t reinvent the wheel
• Move to NoSQL when it really makes sense but
do your best not to administer it

93. Putting all this together
means we should now
easily be able to handle
10+ million users!

94. To infinity…..

95. User >10mil:
• More fine tuning of your application
• More SOA of features/functionality
• Going from Multi-AZ to multi-region
• Potentially needing to start building custom
solutions
• Deep analysis of your whole stack

96. Next Steps?
READ!
• aws.amazon.com/documentation
• aws.amazon.com/architecture
• aws.amazon.com/start-ups

97. Next Steps?
START USING AWS
aws.amazon.com/free

98. Next Steps?
ASK FOR HELP!
• forums.aws.amazon.com
• aws.amazon.com/support
• Your Account Manager
• A Solutions Architect

99. THANKS
FOR
LISTENING!
Mohan Vedula – mohanv@amazon.com
Brett Francis – brettf@amazon.com

100. © 2014 Amazon.com, Inc. and its affiliates. All rights reserved. May not be copied, modified, or distributed in whole or in part without the express consent of Amazon.com, Inc.© 2014 Amazon.com, Inc. and its affiliates. All rights reserved. May not be copied, modified, or distributed in whole or in part without the express consent of Amazon.com, Inc.
Scaling on AWS for the
First 10 Million Users
Mohan Vedula, Enterprise Solutions Architecture
Brett Francis, Enterprise Solutions Architecture
March 26, 2014
Thank you!