The html pages are setup on two redundant virtual servers
in the east coast region of the amazon cloud
that will scale out to multiple load balanced
virtual machines with automatic fail over as the load increases. Round Robin DNS is used to spread the load evenly across the machines. The site is replicated across the country in Nevada on a hot spare connected to a different ISP that can be instantly switched in if the primary sites have problems.
The database information is stored in multiple shards in Simpledb. Physically, its stored redundantly on multiple geographically separate servers with no single point of failure. The database can handle more than 10,000 writes per second. It will auto scale onto multiple machines as load increases.
This image is saved on multiple redundant servers in multiple independant physical facilities with 11 9's of reliability. It is being served from a pool of over 100 webservers at each of the following locations and is served from the closest one to the request for better performance
- United States
- Ashburn, VA
- Dallas/Fort Worth, TX
- Los Angeles, CA
- Miami, FL
- New York, NY
- Newark, NJ
- Palo Alto, CA
- Seattle, WA
- St. Louis, MO
- Hong Kong
As is, the system currently supports peak data transfer speeds of 1,000 megabits per second and peak request rates of 1,000 requests per second. It can be scaled up in 48 hours if this isn't enough. Obviously, we belive in a "make the rubble bounce" level of computer provisioning.