Twitter: We'll Stop Breaking Soon, We Swear

By Niraj Chokshi

Fail Whale -- coletivomambembe-flickr -- photo.png
coletivomambembe/flickr

Twitter wants you to forget the fail whale. The service has struggled to deal with its rapid growth and the problems that arise as a result, but the company hopes that a new data center and some innovative back-end solutions will help hasten updates and reduce downtime.

In a post yesterday, Twitter engineer Jean-Paul Cozzatti explained that a database request locked up the service on Monday, leaving many users unable to sign in or sign up for Twitter. The problem was the latest issue for the rapidly growing service, the maintenance of which is like "building a rocket in mid-flight," Cozzatti wrote.

Building that rocket should become a little easier soon, though. In a separate post yesterday, the company announced that it would be moving into its first custom data center in Utah and plans to open more over the next two years. "
Having our own data center will give us the flexibility to more quickly make adjustments as our infrastructure needs change," Cozzatti wrote.

And, last week, engineer Larry Gadea explained that a new project called Murder -- which is also the collective noun for a group of crows -- was able to reduce the time it took to update Twitter's servers from 40 minutes to 12 seconds. Murder uses the BitTorrent protocol to decentralize updates: rather than relying on one system to distribute an update, each server downloads the update and then starts sharing it with the other servers. Gadea explained the process in-depth in a talk he gave in Montreal (below).

See web-only content:
http://www.theatlantic.com/technology/archive/2010/07/twitter-well-stop-breaking-soon-we-swear/60234/

Twitter - Murder Bittorrent Deploy System from Larry Gadea on Vimeo.


This article available online at:

http://www.theatlantic.com/technology/archive/2010/07/twitter-well-stop-breaking-soon-we-swear/60234/