Main Content

The Gilt Effect: Handling 1000 Shopping Cart Updates per second in Rails

About the Talk

May 5, 2009 6:45 AM

Las Vegas, NV

Las Vegas, NV

Gilt Groupe is a fascinating e-commerce business, where luxury items are sold at a discount in “flash” sales that mimic the New York sample sale experience. In this model, passionate buyers rush to grab items in a time-sensitive shopping cart, choose what they want, and check out within seconds.

These updates would normally require reading and writing to a centralized database, currently PostGres running on Joyent’s Accelerators. Relatively new plugins for threaded database connections, and PostGres’ lack of read/write slaves, pose interesting technical challenges.

Our solution combines a bank of servers at the edge of the network, running on Amazon EC2 with local data stores. An optional queuing layer sits between these servers and the centralized database at Joyent. We’ll discuss our architecture and algorithms to address the problem of ACID shopping cart updates. The solution dynamically scales to handle load, optimizing cost while maintaining a 250MS response time.

Ratings and Recommendations

This Talk hasn't been rated yet. Sign In to rate Talks.

comments powered by Disqus