8. Minimum Viable Product (MVP)
! Release quickly
! Limited core features
! Iterate in production
! Decisions based on data
“Production is truly the only place
you can validate your code”
- Michael Nygard, author of ‘Release it!
15. Define Your MVP’s Building Blocks
Compute
Object Store
Load Balancer
SQL Database
Identity & Access
Data Warehouse
Application Lifecycle
CDN
Recommendation Engine
Social Login SSO
BI Analytics
Web Server
Reverse Proxy
My Second Cool FeatureMy Cool Feature
25. Services are Black Boxes Exposed via APIs
My Cool Feature
Iterate, even re-
write internal
implementation
API is stable, with
few changes,
potentially
versioning
API
e.g JSON / REST
over HTTP
30. Loose Coupling Enables Scale-out and Resiliency
Use Circuit Breakers
Temporarily bypass
unresponsive
service. Switch to
degraded mode
transactions
31. Auto Scale, Load Balance, Monitor, HA Assure
Each Service Separately
32. Statelessness Enables Scale-out
Separate State and Data from Compute Instances
Load Balanced, Auto Scaling
pool of EC2 Workers
Scalable Services for
State and Data
ElastiCacheDynamoDBS3
33. Amazon DynamoDB
Zero Admin NoSQL Service
Unlimited Storage
Provisioned Throughput
<10ms response time
Durable on SSD
34. Cloud-SOA
Service
Oriented
Architecture
! Cloud-native New Applications
! SOA principles leveraging Cloud
services
! Architect for
• Automation
• Scale
• Cost
• Availability
• Data
! Applications are
• Always-on
• Self-healing
• Anti-fragile
35. Implemented by Two-Pizza Teams
! Small, autonomous teams
! Full responsibility for Service X
! Decoupled launch schedules
! No handovers
! Trust
46. « Cloud Instance is an
implementation of a
known, good state »
Dr Rich Wolski, UCSB
47. AMIAMIAMI
YOUR CODE
CORE SERVICES
3rd PARTY
LIBRARIES
OPERATING
SYSTEM
YOUR CODE
CORE SERVICES
3rd PARTY
LIBRARIES
OPERATING
SYSTEM
YOUR CODE
CORE SERVICES
3rd PARTY
LIBRARIES
OPERATING
SYSTEM
Pre-baked Image Base OS Image + Orchestration
58. Go Global – in Minutes
! Launch in a test market
! A/B Testing
! Collect all logs
! Analytics ! Insights ! Iterate
! or Pivot
Once you hit the sweet spot –
! Replicate to all regions
64. « Want to increase innovation?
Lower the cost of failure »
Joi Ito
MIT MediaLabs
65. AWS Lowers the Cost of Failure
Experiment
Often
Fail quickly at
a low cost
More
Innovation
Experiment
Infrequently
Failure is
expensive
Less
Innovation
“ON-PREMISE”