The Facebook-led Open Compute Project is scaling up its efforts to promote open standards for very large data centers, with benefits that should trickle down to cloud computing startups and enterprises.
The Open Compute Project was unveiled in April as a spinoff of Facebook's green data center initiatives. Facebook's initial contributions of server hardware and rack designs were based on those used in its Prineville, Ore., data center, which it says was able to achieve a power usage effectiveness (PUE) ratio of 1.07, indicating only a small fraction of power waste. That compares with 1.5 at other Facebook facilities, which is already in the "best practice" category according to U.S. Environmental Protection Agency standards. A perfect PUE would be 1.0.
While Facebook is taking a leadership role in the project, with its director of hardware and supply chain Frank Frankovsky serving as executive director, the Open Compute Project has established a foundation with a board that includes Andreas von Bechtolsheim, a co-founder of Sun Microsystems with a long history in standards development; Don Duet, head of global technology infrastructure for Goldman Sachs; Mark Roenigk, the COO of Rackspace; and Jason Waxman, general manager of Intel's data center group. The group hosted an Open Compute Summit in New York on Thursday, where one of the featured speakers was James Hamilton, a vice president and distinguished engineer on the Amazon Web Services team.
Hardware vendor contributions are coming from companies like Intel, Dell, Asus, and AMD. Bechtolsheim's latest company, Arista Networks, has also pledged support.
"If you look at the brainpower within this room, it's far greater than the brainpower we have at Facebook alone, or just at Amazon," Frankovsky said. The involvement of Goldman Sachs is also significant because it shows large enterprises have an interest in the same kind of efficiencies, he said.
Facebook will continue to innovate independently on data center design, Frankovsky said. The just-announced Facebook data center in Sweden will be "our greenest yet," drawing the majority of its power from hydroelectric sources, he said.
But while companies like Facebook and other large Web operations have had to learn efficient large scale data center design the hard way, the Open Compute Project will allow younger companies like Zyga "to download the Open Compute specs, start from a base, and then innovate on top" as they get to the point of needing to build their own data centers, Frankovsky said. Meanwhile, data center operator Digital Realty has agreed to incorporate Open Compute designs into facilities for those who prefer to lease, rather than build, data center capacity, he said.
Bechtolsheim said large Internet operations have gotten into the business of designing their own servers and racks out of necessity, because of the "gratuitous differentiation" between blade server and rack designs from different companies. "It would be much better for everyone if there was a standard everyone could use," he said, giving Facebook credit "for not only coming out with the idea but contributing the initial designs, very efficient power-optimized designs, and letting the world use these designs.
"This is a momentous time in history because Facebook has decided to truly open up," Bechtolsheim said. He also said it was important that the Open Compute Alliance was working to cooperate rather than compete with other groups such as the Open Data Center Alliance, which is defining common customer requirements such as standards for cloud security.
Amazon's Hamilton said the need for greater efficiency has always existed, but the rise of data centers that power tens of thousands of servers has brought it to the fore. At Amazon Web Services, the demand for server capacity is growing so fast that "just getting the racks into the buildings and powered is a challenge," he said.
"What's changed is that for a few players in the industry--Facebook, Google, Amazon, Microsoft--have gotten to the scale where you can't afford not to have specialists in every area," Hamilton said, which means they've invested more resources in specialties like server power and cooling efficiency. "Some of those resources are being made available to the industry as a whole, and that will push innovation more rapidly," he said.
We welcome your comments on this topic on our social media channels, or [contact us directly] with questions about the site.