Looking at What Keeps Wikia Running

By Alex Handy

September 15, 2008 —  (Page 1 of 3)
Change may be hard to manage, but what about a system where there are millions of changes every month?

For Wikia, a wiki hosting company founded by Wikipedia creator Jimmy Wales and early Wikimedia board member Angela Beesley, change is the lifeblood of its content. The resulting free wiki hosting service receives around 20 million unique hits a month and hosts some of the most authoritative yet esoteric sources of information on the Internet.

Wikia currently hosts such popular repositories of information as Wookiepedia, the definitive source for information on the “Star Wars” universe, and other user-created wikis covering everything from the Muppets to Madden football games. Behind all of these constantly changing sites is Artur Bergman, vice president of engineering and operations at Wikia.

The Swedish-born Bergman came to the U.S. in 2005 to work at Six Apart, the company behind Movable Type and the TypePad blog hosting services. Before that, he worked for Canon. Along the way, he came to appreciate the cost savings of using open-source stacks with standard hardware for almost all networking duties.

Bergman said that the Wikia stack is entirely open source. The more than 100 systems behind Wikia run on typical LAMP (Linux, MySQL, Apache and PHP) stacks. Systems are typically created using Red Hat’s Cobbler tool and the Puppet open-source configuration framework.

Running on LAMP stacks is nothing unique, but Bergman’s take on routing is. He favors standardized hardware routers, built from scratch.

“We run Vyatta,” Bergman said of his router stack. “They basically do an open-source router stack. Our edge servers are normal Linux machines running Vyatta. They have a community release and a professional release. It’s infinitely cheaper than Cisco.

“We’re pushing text. If I [were] pushing gigabits per second, then perhaps it would not be the right solution. But for us, they do BGP, and we don’t have to worry about routing tables overflowing our memory, because we just put in a 2G RAM stick. It also saves us from a hardware point of view because it runs on normal servers. And I can tcpdump on my router, which is fantastic. No mirror ports need to be set up.”

Related Search Term(s): open source, server management, Wikia

Share this link: http://www.sysmannews.com/link/32832