To increase the power efficiency of its data centers, Facebook decided to build its servers for its Pineville, Ore., data center from scratch.
And it looks like the plan worked. Facebook’s latest data center in Oregon will use 38 percent less energy than the rest of its data centers and will cost 24 percent less to deploy and maintain than the rest of its centers. Facebook seems pretty proud of its accomplishment, too, since it is opening up the specifications of its new data centers and servers to the public and creating an open source project, called Open Compute Project, to build a more efficient data center.
The project is a web site that keeps track of the latest tweaks and modifications Facebook makes to its new hardware specifications. The company wants outside readers to follow the changes and drop in suggestions about improvements to the data centers and hardware — such as suggesting a specific type of fan or power supply. Facebook said it is working with Dell, Hewlett-Packard, Rackspace, Skype and Zynga to develop more efficient servers and data centers through the project.
As part of the project, Facebook has published the specifications of the hardware in its newest data center. It also published the specifications for the data center housing the servers. The company first announced it would build a data center in Oregon in January last year. Facebook intended to make the data center more green by reusing the air to cool servers to heat the office and removing the need for an uninterrupted power supply that would sap additional electricity unnecessarily.
Crowd-sourcing has worked in the past for software projects — ranging from open-source programming projects like Linux to video game development. This is a pretty high-profile attempt at crowd-sourcing the creation of an efficient data center, which might prove to be a good way to offload the costs of having to deploy and test new hardware in-house. That’s assuming it is able to build up a strong community around the project. The landing page has received nearly 700 “likes” since it was announced just this morning.
VB's research team is studying web-personalization... Chime in here, and we’ll share the results.