If you’re a longtime Facebook user, chances are pretty good that you uploaded some photos to the social network that you haven’t actually viewed in years. And you’re not alone: Facebook stores hundreds of billions of photos, with hundreds of millions uploaded every day.
Facebook’s apparently decided on a solution to that immense problem: cold storage in its Prineville, Oregon facility. While the facility is still under construction, the company hopes to have the first of three phases up and running later in 2013, according to The Oregonian. There will be three data hubs, each containing an exabyte, or one quintillion bytes, of data.
Facebook executives and engineers have made no secret of their company’s storage issues. “We have to keep lots of servers around, this costs lots of data,” Jay Parikh, vice president of infrastructure at Facebook, told an audience at the Open Compute Summit earlier this year. When it comes to photos, users also want their images served as quickly as possible, which constrains Facebook’s options in terms of backend infrastructure—in other words, no storing old user photos to tape. We can load that for you in a day or two isn’t exactly something that works for this audience.
Parikh described Facebook’s solution, in which a cold-storage client grabs a particular file and sends it to a staging service, where it’s broken up into chunks; from there, the streaming service works with the storage-device service, which helps place those chunks into the appropriate raw storage nodes. The process is relatively quick, but demands a delicate dance between hardware and software in order to keep everything running in the smoothest and most energy-efficient way possible.
Michael Kirkland, a Facebook communication manager, suggested to The Oregonian that retrieving a photo from cold storage would take seconds or milliseconds. So there you have it: all that engineering expertise and infrastructure, just so you can pull up that embarrassing photo from that party in ’08.