digital workloads on aws
TRANSCRIPT
© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Chris So, Business Development Manager
17 June
New Digital Workloads on AWS Amazon.com, 17,
ZALORA and Capital One
CREDIBLY INNOVATE PHOTO HERE
Case sharing - Innovation TODAY’S AGENDA
1. Amazon.com – Innovation, Omni Channels
2. 17 – Social Media and Live Streaming
3. ZALORA – eCommerce 4. Capital One – Alexa Skill
Get out of bed
Go to AWS event
Learn something
Go back to bed
Digital Disruption
Keys to Digital Transformation Lean Enterprise
• Customer Centric • Minimum Viable Products • Metrics-driven culture • Low-cost experimentation • Failure-embracing culture
IT Enablers
• Microservice architecture • DevOps • Agile – Scrum • CI/CD • Blue-Green and A-B groups • Machine Learning • Internet of Things (IoT) • Mobile • Big Analytics
Enabled By
CLOUD
Can’t Digital Transformation occur on Premise?
• Global economies of scale • Immediate access to latest technology • Improved security and risk visibility • Open – access to industry leading solutions • Elasticity – capacity alignment • Improved resiliency – design for failure • Lower cost of experimentation and innovation • Quicker global expansion • Easier to standardize and automate • Easier to link the supply chain - ecosystems
CLOUD
= 50 million deployments a year
Thousands of teams × Microservice architecture
× Continuous delivery × Multiple environments
~11.6s
Mean time between deployments (weekday)
~1,079
Max number of deployments in a
single hour
~10,000
Mean number of hosts
simultaneously receiving a deployment
~30,000
Max number of hosts
simultaneously receiving a deployment
DEPLOYMENTS AT AMAZON.COM
Deploying More Frequently Lowers Risk
Smaller Effort “Minimized Risk”
Frequent Release Events: “Agile Methodology”
Time
Cha
nge
Rare Release Events: “Waterfall Methodology”
Larger Effort “Increased Risk”
Time
Cha
nge
Monolith development lifecycle
developers
release test build
delivery pipeline app
Service-Oriented Architecture (SOA) Single-purpose primititves Connect only through APIs “Microservices”
Microservice development lifecycle
developers delivery pipelines services
release test build
release test build
release test build
release test build
release test build
release test build
Two-pizza teams Full ownership Full accountability Aligned incentives “DevOps”
It’s Easy for Anyone to Start an Ecommerce Business
This Leads Eventually to Omni Channel Offerings
O2O < > O2O
FromThis…
? … to this
?
FromThis…
… to this
This is by TW Company M2comm
Cloud-based ERP integration
Real-Time Data Push Custom low-power RF
Low Battery Consumption e-Ink displays
Your Life’s Moment 17 –
25
million
streams watched/month
6 million
MAU
12 million
Downloads
What’s the need of 17 architecture?
Scalable Available Personalized
Grow with the users Always there for users Understand the users
The journey of our architecture
User 100 – Launch ASAP
First 100 users
Don’t even think about scalability Launch and verify the idea ASAP
AmazonRoute 53
Amazon EC2 MongoDB
Request
User 100,000 – CDN and Cache
User 100,000
Cache the database Use CDN to deliver the live streaming content
CDN AmazonRoute 53
Amazon EC2
MongoDB Request
Amazon ElastiCache
User 1 million – Design for failure
Design for failure
Failures are the norm, not exceptions
Suppose the rate of failure of one machine is once every 10 years (120 month)
The mean time of failure (MTTF) is 1 month if you have 120 servers
Always assume that things will go WRONG, and design for it
Design for failure
AmazonRoute 53
Amazon EC2
MongoDB
Amazon EC2
Amazon EC2
MongoDB
MongoDB
Elastic Load Balancing
Multi-AZ Multi-AZ
Mix spot and on-demand to save the cost
TIP: Use C3 instance for spot
Amazon ElastiCache
User 5 million – Build loosely coupled systems
Our system was a monolithic system consists of
API Server Streaming Server Worker
Application Server
API Server Streaming Server Worker
Application Server
We discovered a bug the API server didn’t send requests to worker
API Server Streaming Server Worker
Application Server
After fixed, the overloaded worker crashed the whole server
Split the service, so that it’s easier to scale, and fail independently
Build loosely coupled systems
API Server
Streaming Server
Worker
API Server
API Server
Streaming Server
Worker
Worker
AmazonSQS
API Cluster Worker Cluster Streaming Cluster
Monitor for each service, and design for failure
4M customers
7M visits weekly
11M mobile app download
2M orders shipped in Q4 2015
Before • Platform fully hosted in a physical
DC in Hong Kong • Average capacity utilization 10% • Living dangerously during peak • Lead-time for adding peak
capacity – 10 days
ZALORA cloud journey
After • “Double 12”: China's Cyber
Monday 6X • Redshift for click stream analysis • Live migration in 30 days for all 8
sites • Average 40% utilization; • 4X capacity increase in 15 mins • 1-2s faster page loads in ID & PH,
~10% better CR
6X
6X ?
visits
resources
Amazon Route 53
User Amazon
CloudFront
Magento
RDS Master (Multi-AZ)
Elastic Load
Balancer
RDS Slave (Multi-AZ)
Magento
Availability Zone Availability Zone
Backup Storage
Static website *.html, *.js *.css
*.jpg *.mp4
S3
Redis Read Replica
Redis Master
RDS Read Replica
RDS Read Replica
Magento Magento
Magento Admin
Private subnet Office
VPN
A Scalable Platform
Stock Exchange NASDAQ and NYSE
1000+ Skills
Alexa Skills Kit architecture
Amazon Alexa
service
Developer’s application
service
Amazon’s Developer
Portal Application, intents, sample data, developer service URL endpoint
Configured through portal
User intents and arguments are sent to the developer service
GUI cards are rendered in the Amazon Alexa app
User audio is streamed to the service
Audio responses are rendered on-device
Text response and/or GUI card data is returned
Alexa is always learning. Alexa gets smarter by learning new skills. Developers can create new skills for Alexa.
Alexa is ALWAYS LEARNING
Capital One Skill
Capital One’s Alexa approach June: A few developers buy Echos
July: Full day tech offsite & side of desk project kickoff
August: Rapid prototyping and expanding Capital One skill
Goal: Pair Alexa with the Capital One app and
allow users to get their credit card balance
March 2016: Official Launch
Capital One skills focus
Read-only information Transactional skills Experimenting
• Default accounts (credit card, bank, loans)
• Account balances • Bill due date • Last payment • Last transactions • Interest rate
• Pay bill(s) • Transfer $
• App usage Patterns • O-Auth • Customer service/
support • Customer
acquisition • Alexa adoption • Alexa evolution
Skill development segmented into three priority buckets
Responding requires a new model
Focus on differentiating your company
Innovate at start-up like speed
Reduce risk
Thank You