An Open Compute Primer
Beginning today, the third Open Compute Summit will be held in San Antonio at the headquarters of Rackspace Hosting. Started little over a year ago, the Open Compute Project (OCP) aims to change how the hardware that powers scale-computing technology is built. But what exactly is OCP and what are some of the benefits? Here is a quick primer.
Origins of Open Compute
Facebook started the OCP as they looked to scale their datacenter infrastructure in the most efficient and economical way possible. In addition to the actual hardware, one of the largest datacenter costs involves powering the servers.
By building an efficient server, not only do companies save on power costs, they can also reduce their carbon footprint. At the 2012 Summit Keynote saying, Lew Moorman, President of Rackspace noted, “we need lower cost, reliable and greener hardware and datacenters. That is what is happening. The Open Compute Project is real.”
Standardization In The Open
During the keynote, Frank Frankovsky, Director of Hardware Design and Supply Chain at Facebook, described the typical way that hardware is made for scale computing companies. “When suppliers approach all of us, they get different requirements. This leads to a whole bunch of unique designs. The amount of innovation wood they can put behind each of these products is relatively small.”
Frankovsky said the only standard for datacenters is that the racks holding the servers should be 19” wide without a height, depth or mounting scheme. In fact, this standard is quite archaic with origins back to railroad signaling relays. “Because we’ve followed that standard for so long, we’ve ended up what I call ‘racks gone bad.’ Not only is it unsightly, it impacts service at scale,” Frankovsky says. These ‘racks gone bad’ can be anything from oversized servers being jammed into a shallow depth rack to huge power cables that are difficult to work with.
Continued on the next page
Follow Technorati