Read more on "Server Hardware Spotlight" »

Rivals Google, Facebook Team Up on Efficient Rack Design

Google and Facebook are fierce rivals with a common challenge: how to fit as much processing power into their data centers as possible to meet the always-growing demand for more capacity while keeping down the power and operational costs that come with adding servers.

To address the issue, the two hyperscale players have looked beyond the ongoing competition for online advertising to work together on a new standard that could help each company in the data center. Google and Facebook worked together in the Open Compute Project (OCP) to develop what they're calling the Open Rack v2.0 standard, which officials with both said will increase the performance and efficiency in their respective facilities.

The two companies will present the standard at the OCP Engineering Workshop Aug. 10 at the University in New Hampshire with hope of getting it approved.

The new standard comes five months after Google joined the OCP, which Facebook launched in 2011 in hopes of creating open-source standards for highly energy-efficient data centers and IT hardware. Like other Web 2.0 companies—such as Google, Microsoft, eBay and Amazon—the social networking giant needed to find data center hardware that was more efficient than the major OEMs were selling, and turned to in-house development. Facebook engineers spent two years developing new hardware designs that were 38 percent more efficient and 24 percent less expensive to run than the systems they had been running.

The OPC has since grown to more than 160 member organizations—from tech vendors (including major systems OEMs like Hewlett Packard Enterprise, IBM, Cisco Systems and Dell) to end users—and encompasses all areas of the data center. Many system makers are adopting open designs in their hardware offerings.

Since joining the OCP in March, Google has worked with Facebook on the Open Rack Standard, which encompasses much of what Google engineers have learned over the past several years in developing their own in-house systems that are smaller and more efficient, enabling the company to put more of them into smaller footprints.

A key contribution Google made to the OCP Open Rack standard was bringing the 48-volt power technology it had developed internally. The previous version of the standard had used a 12-volt technology, but Google officials said they had seen significant benefits by using the 48-volt technology in their own data centers for six years.

"Google developed a 48V ecosystem with payloads utilizing 48V to Point-of-Load technology and has extensively deployed these high-efficiency, high-availability systems since 2010," Debosmita Das, technical program manager, and Mike Lau, technical lead manager at Google, wrote in a post on the company blog. "We have seen significant reduction in losses and increased efficiency compared to 12V solutions. The improved SPUE [system power usage efficiency] with 48V has saved Google millions of dollars and kilowatt hours."

In a blog post announcing Google's becoming an OCP member, John Zipfel, technical program manager at Google, wrote that as "the industry's working to solve these same problems and dealing with higher-power workloads, such as GPUs for machine learning, it makes sense to standardize this new design by working with OCP. We believe this will help everyone adopt this next-generation power architecture, and realize the same power efficiency and cost benefits as Google."

According to Das and Lau, the Open Rack v2.0 standard specifies a 48-volt power architecture with a modular, shallow-depth form factor that will enable users to put high-density OCP-based racks into their data centers. If the standard is adopted, it can begin to be implemented into system makers' designs.


Originally published on eWeek.
This article was originally published on August 8, 2016
Page 1 of 1

Read more on "Server Hardware Spotlight" »
Thanks for your registration, follow us on our social networks to keep up-to-date