When Facebook decided to open-source the hardware that makes its data centers some of the most energy-efficient and innovative in the tech industry, others tech giants were eager to join in and open-source their Internet hardware as well. But not the biggest do-gooder, greenster, data center magnate of them all: Google.
The goal was to drastically and rapidly improve energy efficiency in data centers. This coalition of web and hardware giants is known as the Open Compute Project, and so far, only one name in mainstream tech is still glaringly absent from its membership roster.
While Facebook and Google have a historically tense relationship over business matters large and small, no one involved in Open Compute can put a finger on exactly why Google refuses to get involved. After all, the project fits right in with Google’s “do no evil” ethos and campaigns for more ecologically friendly data centers.
And while Google has previously hinted that energy efficiency is part of its competitive advantage, many other companies that could say the same thing have still wholeheartedly devoted their time and intellectual property to Open Compute.
So what’s the hang-up with Google?
VentureBeat has reached out to Google several times over the past year; until last week, no one has been able to speak directly about Google’s data center innovations or its official stance on the Open Compute project.
When we finally got the chance to sit down with Urs Hölzle, Google’s infrastructure czar, he confirmed what third parties have long been telling us: Google won’t play ball because it thinks its hardware constitutes a competitive advantage. (We’ll be publishing Hölzle’s fuller thoughts on the subject tomorrow morning.)
In a recent conversation at Facebook’s Menlo Park campus, hardware design director Frank Frankovsky, the guru behind much of Facebook’s Open Compute work, highlighted just how many companies outside Facebook have piled onto the Open Compute bandwagon.
“There have been a number of new contributions,” he said. “An example would be the Financial District, one of the earliest adopters of Linux. … They’ve been pretty passionate about Open Compute, and they kicked off projects with Intel and AMD [to build and customize their own hardware].”
In the past six months alone, newcomers like AMD, HP, West Digital, Fidelity, Salseforce, Applied Micto, Quanta, NTT Data, ZT Systems, Emulex, DataDirect, Tencent, and Vantage data centers have joined Open Compute, which already counts Rackspace, Intel, Salesforce, Alibaba, and many others as members.
Some companies have even turned over their own projects, such as AMD’s Roadrunner, Tencent’s Project Scorpio, and Intel’s Decathelete, to Open Compute for its incubation committee or for inclusion in the project.
“This reminds me a little bit of the Brady Bunch,” said Frankovsky. As the various projects are converged and standardized, they become more valuable for large enterprises and manufacturers as well as for the project’s members themselves.
“Open Compute might evolve into a broader focus over time,” he said. There’s a possibility that the same open-source workflow that’s being applied to this data center hardware might someday be applied to other networked devices and even mobile devices. Can you imagine, for example, how the community might work together to solve issues like cell phone battery life?
“We might spread our wings,” Frankovsky concluded, “but for now, we want to stay focused on data centers, servers, and storage.”
And for the companies that have so far been involved in Open Compute, the Facebook team and Open Compute partners we’ve spoken to have had nothing but positive comments on the experience — and on the project’s impact on their collective bottom line.
“The benefits of sharing so far outweight the beneifts of keeping it all closed,” said Frankovsky. Of Google’s “competitive advantage” argument, he added, “I have trouble getting myself in the mindset of thinking the infrastructure is a strategic advantage from a cost perspective. It’s really more about serving the end users — that’s what differentiates a business.”
Frankovsky and the rest of the Facebookers we talked to during this visit never mentioned Google by name, but this was the competitor clearly on their minds. From the Open Compute perspective, Google’s true competitive advantage is in its search algorithm, its web-based software, its colossal reach, and not in its racks.
When we brought up the search company directly, Frankovsky said, “We’d love to see a lot more of our peers and competitors show up for this. I know they’ve made mistakes we’re about to make. … By sharing those great successes as well as the mistakes, we could save a lot of time and effort.”
“We keep reaching out,” he said. “Sometimes they show up in listen-only mode.”
Frankovsky’s “missed connections” statement seems resigned but not angry. However, other Facebookers we’ve spoken to, including data center employees, have not been shy about expressing their frustration over Google’s recalcitrance.
“This isn’t an all or nothing approach,” said Mike Schroepfer, Facebook’s engineering VP. He said that, if it so chose, Google could open-source bits and pieces of its data center and server technology without revealing the full recipe of its secret sauce.
“It’s disappointing when people don’t participate when they can selectively share with the community,” he said. “There’s lots of room for how you run the software to create a competitive advantage.”
Ultimately, he said, he expects Open Compute’s innovations to achieve a Linux-like stature and ubiquity. “The investment of the industry in Linux caused it to be superior to [Sun Microsystems’] Solaris over time,” he said. “It’s hard to bet against a large collective of people applying all their IP in a project.”
Schroepfer, who ran Firefox development for many years while working for Mozilla, has made his career in open-source and said that open source has become “the de facto way you do software.” But, he added, it’s been harder to get even seasoned open-source advocates like Google onboard when it comes to hardware.
“[It’s essential] for everyone in the industry to get as power-efficient as possible, both for their bottom line and for the planet,” he said.
And he thinks Open Compute is well on its way to accomplishing that goal. “It’s astonishing to see the quality of [Open Compute] technology in production,” he said, in reference to a visit he made to Facebook’s Prineville data center, where he was making repairs to a Freedom open-source server.
Open Compute isn’t a perfect project, and its players aren’t perfect, either. But we can’t shake the sense of disappointment that Google hasn’t yet stepped up, with all its knowledge about energy efficiency and its many years of experience in building and optimizing data centers, to participate in the project.
But this isn’t the first Facebook-led open-source project Google’s backed away from. For comparison, we bring up Ringmark and the W3C’s Core Mobile Web Platform Community Group. These Facebook-fronted efforts are intended to fix one of the worst pain points in technology — and one of the points of greatest promise — the mobile web.
But as Facebook stepped forward to lead the charge and open-source its information, Google turned its back on the project, an especially odd move given Google’s interest in the mobile ecosystem vis-à-vis Android and Chrome’s new mobile browser, not to mention Google’s historical and significant participation in the free and open-source software movement. At that time, a Google spokesperson declined to speak directly about the W3C group, instead focusing on the company’s own mobile browser and operating system work.
It’s a bit naïve, a bit “kumbaya,” to expect two titans viciously warring for revenue to come to a round table and collaborate on issues like innovative mobile tech and greener data centers. But somehow, without striking a blow, Facebook has managed to give Google a black eye. Perhaps, if only for PR purposes, it’s time for Google to get its “kumbaya” back.
Images of Facebook’s Prineville data center courtesy of Jolie O’Dell
The audio problem: Learn how new cloud-based API solutions are solving imperfect, frustrating audio in video conferences. Access here