Video: Facebook’s ultra-efficient server, a guided tour

I’ve been spending the afternoon at the SC11 supercomputer conference, taking place this week in Seattle, where one of the more unexpected companies on the show floor is Facebook.

Actually, it won’t be a surprise if you’re familiar with the Open Compute Project, an open-source-style project that started with the social network’s efforts to share with the industry what it learned in the process of creating efficient data centers, such as its location in Prineville, Ore.  The Open Compute project recently announced the creation of a foundation to lead the initiative.

So what makes these servers different? Amir Michael, a Facebook hardware design manager, gave us a guided tour at the SC11 conference this afternoon, in the video above.

  • http://chuck.goolsbee.org chuck goolsbee

    These servers are a breeze to work on, saving countless hours in repair & upgrade cycle time. For example, performing a RAM upgrade on an Open Compute Server takes about a minute max, whereas on a traditional 1U server can take 10+ minutes to un-rack, disassemble, swap DIMMs, reassemble, and re-rack. Multiply that across thousands of servers and start calculating your time saved! Wonderful machines!