This document provides guidance on scaling a web application from 1 user to over 10 million users on AWS. It recommends starting simply with a single EC2 instance and Route 53, then adding redundancy with multiple instances, load balancing, and SQL databases. As users grow over 1,000 techniques like caching, NoSQL, and auto scaling are introduced. Above 500,000 users more services are split out and automated. Reaching over 1 million requires database sharding or federation. The key strategies emphasized are redundancy, automation, splitting services, and leveraging managed AWS services over custom solutions.

1. © 2015, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
John Pignata, Startup Solutions Architect
August 14, 2017
Scaling Up to Your
First 10 Million Users

2. http://i.telegraph.co.uk/multimedia/archive/02674/CLIMBER_2674482b.jpg

4. Now that’s a lot
of things to read!
This is NOT
where we
want to start!

5. It’s not the single thing that
fixes everything.

6. What do we need first?

7. Some basics…

8. AWS global infrastructure
Region, # of zones
Coming Soon!

9. AWS global infrastructure

10. So let’s start from…

11. You

12. 1 User
Amazon
EC2
instance
Elastic IP
User
Amazon
Route 53

13. “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
c4.8xlarge
m4.2xlarge
t2.micro

14. “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
c4.8xlarge
m4.2xlarge
t2.micro

15. 1 User
• No failover
• No redundancy
• Too many eggs in one
basket
EC2
Instance
Elastic IP
User
Amazon
Route 53

16. Users >1

17. Users > 1
Web
Instance
Database
Instance
Elastic IP
User
Amazon
Route 53

18. Self-managed Fully managed
Amazon EC2 Amazon
DynamoDB
Amazon RDS Amazon
Redshift
Database options

19. • MySQL or Postgres compatible
• Automatic storage scaling (up to 64 TB)
• Up to 15 read replicas
• Continuous (incremental) backups to
Amazon S3
• 6-way replication across 3 zones
Amazon Aurora

20. To NoSQL, or not to NoSQL?

21. Start with SQL databases

22. Why start with SQL?
• Established and well-worn technology.
• Lots of existing code, communities, books, and tools.
• You aren’t going to break SQL DBs in your first millions
of users. No, really, you won’t.*
• Clear patterns to scalability.
*Unless you are doing something SUPER peculiar with the data or you have MASSIVE amounts of it.
…but even then SQL will have a place in your stack.

23. AH HA! You said
“massive
amounts”

24. > 5 TB in year one?
Incredibly data intensive workload?
OK!
You might need NoSQL.

25. Why else might you need NoSQL?
• Super low-latency applications
• Metadata-driven datasets
• Highly nonrelational data
• Need schema-less data constructs*
• Rapid ingest of data (thousands of records/sec)
• Massive amounts of data (again, in the TB range)
*Need!= “It’s easier to do dev without schemas”

26. Users >100

27. Users >100
Web
instance
Elastic IP
RDS DB
instance
User
Amazon
Route 53

28. Users >1000

29. Users >1000
Web
Instance
RDS DB Instance
Active (Multi-AZ)
Availability Zone Availability Zone
Web
Instance
RDS DB Instance
Standby (Multi-AZ)
Load
balancer
User
Amazon
Route 53

30. Elastic Load Balancing
• Highly available
• 1–65535
• Health checks
• Session stickiness
• Monitoring/logging
• Secure Sockets Layer

31. Application Load Balancer
• Highly available
• 1–65535
• Health checks
• Session stickiness
• Monitoring/logging
• Content-based routing
• Container-based apps
• WebSockets
• HTTP/2

32. Horizontally
Vertically

33. Users >100,000

34. Users > 10,000s–100,000s
RDS DB Instance
Active (Multi-AZ)
Availability Zone Availability Zone
RDS DB Instance
Standby (Multi-AZ)
Load
balancer
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 53User

35. What about
performance and efficiency?

36. RDS DB Instance
Active (Multi-AZ)
Availability Zone
Load
balancer
Amazon S3
Amazon
CloudFront
Amazon
Route 53
User
Shift some load around
Web Instances
Availability Zone
Web Instances
RDS DB Instance
Standby (Multi-AZ)

37. RDS DB Instance
Active (Multi-AZ)
Availability Zone
Load
balancer
Amazon S3
Amazon
CloudFront
Amazon
Route 53
User
Shift some load around
Web Instances

38. • Object-based storage
• Highly durable
• Great for static assets
• “Infinitely scalable”
• Objects up to 5 TB in
size
• Optional encryption
Amazon S3

39. • Cache content for faster
delivery
• Lower load on origin
• Dynamic and static
content
• Streaming video
• Custom SSL certificates
• Low TTLs (as short as 0
seconds)
• Free origin fetches?
• Optimized for AWS
Amazon CloudFront

40. Shift some load around
RDS DB Instance
Active (Multi-AZ)
Availability Zone
Load
balancer
Amazon S3
Amazon
CloudFrontUser
ElastiCache DynamoDB
Web Instances
Amazon
Route 53

41. Amazon CloudFront
ResponseTime
ServerLoad
Response
Time
Server
Load
Response
Time
Server
Load
No CDN CDN for Static
Content
CDN for Static &
Dynamic Content
0
20
40
60
80
8:00AM
9:00AM
10:00AM
11:00AM
12:00PM
1:00PM
2:00PM
3:00PM
4:00PM
5:00PM
6:00PM
7:00PM
8:00PM
9:00PM
VolumeofData
Delivered(Gbps)

42. Shift some load around
RDS DB Instance
Active (Multi-AZ)
Availability Zone
Load
balancer
Amazon S3
Amazon
CloudFront
Amazon
Route 53
User
ElastiCache DynamoDB
Web Instances

43. Amazon DynamoDB
• Managed NoSQL database
• Provisioned throughput
• Fast, predictable performance
• Fully distributed, fault tolerant
• JSON support
• Items up to 400 KB

44. Amazon ElastiCache
• Managed Memcached or Redis
• Scale from one to many nodes
• Self-healing (replaces dead instance)
• Single digit ms speeds (usually)
• Local to a single AZ for Memcached
• Multi-AZ possible with Redis

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

46. Auto Scaling!

47. Automatic resizing of compute clusters
Define min/max pool sizes
CloudWatch metrics drive scaling
On-Demand or Spot instances
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, us-west-2b
Auto Scaling

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

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

50. November
November traffic to Amazon.com

51. Provisioned capacity
November
November traffic to Amazon.com

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

53. November traffic to Amazon.com
November

54. Auto Scaling
lets you do this!

55. Users > 500,000+
Availability Zone
Amazon
Route 53
User
Amazon S3
Amazon
CloudFront
Availability Zone
Load
balancer
DynamoDB
RDS 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)

56. Users > 500,000+
Availability Zone
Amazon
Route 53
User
Amazon S3
Amazon
CloudFront
Availability Zone
Load
balancer
DynamoDB
RDS 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)

57. Use automation

58. AWS application management solutions
Convenience Control
Higher-level services Do it yourself
AWS
Elastic Beanstalk
AWS
OpsWorks
AWS
CloudFormation
Amazon EC2

59. AWS CodeDeploy
• Deploys your code to a “fleet” of EC2 instances
• 1–10,000s of instances
• Automatically schedules updates (multiple zones)
• Application and deployment groups described in
YAML-formatted files
• Can reference Auto Scaling groups
• AWS Management Console, AWS CLI, or APIs
• Can be used with Chef recipes or Puppet scripts

60. Users >500,000+
• Monitoring, metrics, and logging
• If you can’t build it internally,
outsource it! (third-party SaaS)
• What are customers saying?
• Try to squeeze as much performance
out of each service/component

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

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

63. SOA
What does this mean?

65. Now that’s a lot
of things to read!
This is NOT
where we
want to start!

66. This is NOT
where we
want to start!
This IS where
we want to start!
Now that’s a lot
of things to read!

67. SOAing
Move services into their own tiers
• Treat them separately
• Scale them independently
It offers flexibility and greater
understanding of each component

68. Loose coupling + SOA = winning
DON’T REINVENT THE WHEEL
• Email
• Queuing
• Transcoding
• Search
• Databases
• Monitoring
• Metrics
• Logging
• Compute
Amazon
CloudSearch
Amazon SQSAmazon SNS
Amazon Elastic
Transcoder
Amazon SWFAmazon SES
AWS Lambda

69. • Reliable (Multi-AZ)
• Scalable (unlimited messages)
• Secure (queue authentication)
• Simple (simple APIs)
Application services – Amazon SQS
SQS
messages
Get
message
Instance
Put
message
Instance
Amazon SNS Topic
Publish
notification
Queue is subscribed
to topic

70. Compute/platform – AWS Lambda
• Functions triggered by events
• JavaScript, Java, Python, and C#
• Managed
• Implicit scaling
S3 bucket
Lambda
Push: event
notification
DynamoDB
Pull: DynamoDB
Stream
Amazon
Kinesis
Pull:
Amazon Kinesis
stream

71. 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
• Don’t build your own!
S3 bucket
Lambda
Push: event
notification
DynamoDB
Pull: DynamoDB
Stream
Amazon
Kinesis
SQS
messages
Get
message
Instance
Put
message
Instance
Amazon SNS Topic
Publish
notification
Queue is subscribed
to topic
Pull:
Amazon Kinesis
stream

72. Users >1,000,000

73. Users >1 million+
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 (SOA)
• Serving content smartly (Amazon S3/CloudFront)
• Caching off DB
• Moving state off tiers that auto scale

74. Users >1 million+
RDS DB Instance
Active (Multi-AZ)
Availability Zone
Load
balancer
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
DynamoDB
Amazon SQS
ElastiCache
Worker
Instance
Worker
Instance
Amazon
CloudWatch
Internal App
Instance
Internal App
Instance Amazon SES
Lambda

75. The next big steps

76. Users >5 million–10 million
Database issues?
How can you solve?
• Federation – splitting into multiple DBs based on function
• Sharding – splitting one dataset up across multiple hosts
• Moving some functionality to other types of DBs (NoSQL, Graph)

77. Database federation
• Split up databases by function/purpose
• Harder to do cross-function queries
• Essentially delays sharding/NoSQL
• Won’t help with single huge functions/tables
Forums DB
Users DB
Products
DB

78. Sharded horizontal scaling
• More complex at the application layer
• No practical limit on scalability
• Operation complexity/sophistication
• Shard by function or key space
• RDBMS or NoSQL
User ShardID
002345 A
002346 B
002347 C
002348 B
002349 A
CBA

79. Shifting functionality to NoSQL
• Similar in a sense to federation
• NoSQL vs. SQL
• Leverage managed services like
DynamoDB
Some use cases:
• Leaderboards/scoring
• Rapid ingest of clickstream/log data
• Temporary data needs (cart data)
• “Hot” tables
• Metadata/lookup tables
DynamoDB

80. A quick review

81. A quick review
• Multi-AZ your infrastructure.
• Make use of self-scaling services – ELB, Amazon S3,
Amazon SNS, Amazon SQS, Amazon SWF, Amazon SES,
etc.
• Build in redundancy at every level.
• Start with SQL. Seriously.
• Cache data both inside and outside your infrastructure.
• Use automation tools in your infrastructure.

82. A quick review continued
• Make sure you have good metrics/monitoring/logging
• Split tiers into individual services (SOA)
• Use Auto Scaling once you’re ready for it
• Don’t reinvent the wheel
• Move to NoSQL if and when it makes sense

83. 10+ million users!

84. To infinity...

85. User >10 million
• More fine-tuning of your application
• More SOA of features/functionality
• Going from Multi-AZ to multi-region
• Possibly start to build custom solutions
• Deep analysis of your entire stack
• Amazon EC2 Container Service
• AWS Lambda

86. Next steps?
READ!
aws.amazon.com/documentation
aws.amazon.com/architecture
aws.amazon.com/start-ups
START USING AWS:
aws.amazon.com/free/

87. Ask for help!
forums.aws.amazon.com
aws.amazon.com/premiumsupport/
Your account manager
A solutions architect

88. Thank you!