3. Our current environment
•Thousands of servers
•Petabytes of storage
•Multiple data centres
•80% virtualised with VMWare
•Committed to the cloud and reducing on-premises
infrastructure
4. Why AWS?
• Unpredictable demand for websites
• True utility pricing
• Want to bring up services quickly
• Reputation
• New management have worked with it before
5. Experiences getting started
• Developers and sys admin had no experience
– Put in place intense training program
• Change in development practices
– (tightly coupled middleware and code previously)
• Needed new service management process
• New security model
7. Benefits of AWS for WordNerd
• Would have cost us around £25,000 to buy in computing resources
and weeks lead time, or $75.40 per day if using existing kit
• Project cost us $27.60 per day in Amazon resources
• Scaled to take the load as needed
• No manual intervention to scale up and down
• RDS makes databases easier – less management
9. Amazon technology we use
• EC2 (compute)
• EBS (disk)
• RDS (database)
• CloudWatch (monitoring)
• SNS (notification)
• ELB (load balancing)
• Elastic MapReduce (data analysis/HPC)
• VPC (linking back to out site)
10. Challenges
• Rethink of system design
– Design for failure of any component and region
– High performance comes from scale out, not up
– Automate as much as possible (Puppet linked to version control)
• Finding good support partners
• Having to build a new monitoring/alerting system
• Solaris cannot easily be used in Amazon
11. How can Amazon improve
• 64 bit small instances so we don’t need two images
• True cloud database similar to Xeround
• Replicating file system between regions
• Continue to push upwards into PaaS/SaaS