View
596
Download
1
Embed Size (px)
DESCRIPTION
Citation preview
Building Applications
with AWS
Paul Duffy
Empower start-ups
On-demand, no upfront investment, low cost
Cost efficiencies are the sizzle.
Agility is the steak.
Empower SMBs and Enterprises
On-demand, no upfront investment, low cost
100,000s of customers
Across 190+ countries
Full spectrum of customers
From Airbnb to Shell. From Playfish to NASA.
Full spectrum of applications
Web apps and beyond…
Full spectrum of applications
Web apps and beyond…
Spectrum of websites
Static. Dynamic.
Simple. Complex.
Niche. Mass market.
Web applications
Popular use case for AWS
Availability is important
Redundant HA systems are costly and complex.
Capacity is important
“Success failure” as customers flock to a site.
Huge cap-ex required. High risk.
Latency is important
Restricted download speeds, high latency when
deployed from a single geographic location.
Utility computing to the rescue
Elastic infrastructure allows high availability by default.
Scale is there when you need it, without large upfront costs.
Unable to
serve
constituents
Infrastructure
Cost $
time
Large
Capital
Expenditure
Predicted
Demand
Traditional
Hardware
Actual
Demand
Automated
Virtualization
Reach and support a global audience
CloudFront content delivery network.
Latency aware routing for Route 53.
Elastic infrastructure
Decoupled. Stateless. Horizontal scale.
Load balancer
Application servers
Data store
Load balancer
Application servers
Data store
Load balancer
Application servers
Data store
Load balancer
Application servers
Data store
Design for availability
“Everything fails, all the time”
Design for failure. Compute as an interchangeable resource.
Load balancer
Application servers
Data store
Load balancer
Application servers
Data store
Load balancer
Application servers
Data store
Load balancer
Application servers
Data store
Load balancer
Application servers
Data store
Load balancer
Application servers
Data store
Design for automation
Auto-scaling for scale and availability.
Load balancer
Application servers
Data store
Load balancer
Application servers
Data store
Load balancer
Application servers
Data store
Design for automation
Automated provisioning and management.
Design for scale
Follow architecture best practices for
availability and automation
Design for scale
Employ scalable services
Load balancer
Application servers
Data store
Load balancer
Application servers
Data store
Load balancer
Application servers
Data store
Load balancer
Application servers
Data store
Load balancer
Application servers
Data store
Beyond availability and scale…
Sophisticated web applications
Powered by productive tools and services
Managed. Zero admin.
Up and running for production in minutes.
Amazon DynamoDB
Managed NoSQL database service
Consistent, single digit millisecond latencies
Flexible data model
Web apps at all scales
Casual games, social networks, user generated content
Amazon Simple Workflow
Orchestration for distributed workflows
Business process. Image/transcoding processing. Deployment.
Integrated web applications
Public and private subnets across a Virtual Private Cloud.
Full spectrum of applications
Web apps and beyond…
1. Mobile applications
Flipboard, Live TalkBack
2. Analytics applications
Recommendations to click stream analysis
Daily batch processing requirement:
Click stream data (TB’s / day)
Daily online ad spend analysis
Compile Results
Optimize next day’s ad spend
3.5 billion records 71 million unique cookies 1.7 million targeted ads per day Several TBs of Clickstream logs a day
Targeted
Home theater system purchase and is searching for video games (1.7 Million per day)
Traditional infrastructure AWS Cloud
SAN storage 30 servers for compute 3 high-end SQL servers
Cloud Services Hadoop and Cascading “Ad Serving” Integration
Upfront CapEx: ~$500K Recurring OpEx: significant Procurement time: 2 mos. Processing time: 2 days /Job
Upfront CapEx: $0 Recurring OpEx: $13K/mo. Procurement time: 0 Processing time: 8 hours / Job
Clickstream analysis from similar purchase behavior
Analyze Targeted advert displayed
500%
Return on advertising spend.
3. Backup and Disaster Recovery
Durable storage. Physical separation.
Elastic infrastructure for rapid recovery.
High Availability
Backup Storage
Disaster Recovery
DR for regulated
industries
Oracle E-business
Suite DR of dev/test
Enterprise
application backup
and DR
Hot site DR for
enterprise clients
Fast, secure, cost effective backup
and DR for Oracle apps
IT and Operations are more
responsive to the business
Reducing cost of vaulting and DR
implementations by up to 50%
Avoiding $1M of infrastructure and
admin costs
3. Backup and Disaster Recovery
Backup to AWS S3. Recover on-premise.
Recover to EC2.
Pilot Light
User or system
Web
Server
Application
Server
Database
Server
Data
Volume
Web
Server
Application
Server
Database
Server
Data
Volume
Data Mirroring/
Replication
Not Running
Smaller Instance
Amazon Route 53
Pilot Light
User or system
Web
Server
Database
Server
Data
Volume
Web
Server
Application
Server
Database
Server
Data
Volume
Not Running
Smaller Instance
Amazon Route 53
Web
Server
Application
Server
Database
Server Data Mirroring/
Replication
Application
Server
Web
Server
Pilot Light
User or system
Web
Server
Database
Server
Data
Volume
Database
Server
Data
Volume
Start in minutes
Resize as desired
Amazon Route 53
Web
Server
Application
Server
Database
Server Data Mirroring/
Replication
4. Corporate Applications
Amazon’s mission critical corporate applications
Planned migration
Microsoft SharePoint 2010 &
SQL Server 2008
Running within the Virtual Private Cloud.
Microsoft License Mobility
With Software Assurance
Worked with AWS just as with any vendor
Premium support. Solutions Architecture support.
Infrastructure costs down 22%
Procurement time down from over 4 weeks to minutes.
5. Applications on Mars
NASA and the Mars Rover
Left image
Warp
Right image
Warp
Stereo match
3D mapping
Tile
Full spectrum of applications
Web apps and beyond…
Web applications
Availability. Automation. Latency. Scale.
And beyond…
Mobile. Analytics. DR and backup. Corporate apps.
Interplanetary.
Thank you!