How the Open Compute Project helped Facebook reduce costs
Facebook is preparing to open up a 290,000 square-foot data center in Sweden that will exclusively utilize servers designed in-house, reports Ars Technica.
Just as maintaining its own data center is cheaper than relying on cloud providers, so is cutting out traditional server vendors, according to Frank Frankovsky, VP of hardware design and supply chain operations at Facebook. The Open Compute Project was started by the company two years ago.
The servers designed by Facebook (NASDAQ: FB) are known as ODMs, or original design manufacturers, and were built in Taiwan and China. They help the company save on what Frankovsky terms "gratuitous differentiation"--or unique hardware features incorporated by server vendors such as HP (NYSE: HPQ) or Dell, but which do not benefit Facebook.
One example highlighted by Frankovsky was the plastic bezel with the brand logo of a vendor. A study cited by Frankovsky has shown that a "standard" 1U OEM server consumes 28 watts of fan power compared to just 3 watts used by the Open Compute server.
Facebook has no use for the proprietary server management features added by individual vendors.
In total, Facebook says it achieved a savings of 24 percent from its use of this infrastructure, while the lower costs of operating its own equipment helped it cut its ongoing operating costs by another 38 percent. For more on how the Open Compute is making a difference for Facebook, read the full Ars Technica article here.
- check out this article at Ars Technica