If you invented something cheaper, more efficient, and more powerful than what came before, you might want to keep the recipe a closely guarded secret. Yet Facebook took the opposite approach after opening a 147,000-square-foot computing center in rural Oregon this April. It published blueprints for everything from the power supplies of its computers to the super-efficient cooling system of the building. Other companies are now cherry-picking ideas from those designs to cut the costs of building similar facilities for cloud computing.
The Open Compute Project, as the effort to open-source the technology in Facebook’s vast data center is known, may sound altruistic. But it is an attempt to manipulate the market for large-scale computing infrastructure in Facebook’s favor. The company hopes to encourage hardware suppliers to adopt its designs widely, which could in turn drive down the cost of the sever computers that deal with the growing mountain of photos and messages posted by its 750 million users. Just six months after the project’s debut, there are signs that the strategy is working and that it will lower the costs of building—and hence using—cloud computing infrastructure for other businesses, too.
Facebook’s peers, such as Google and Amazon, maintain a tight silence about how they built the cloud infrastructure that underpins their businesses. But that stifles the flow of ideas needed to make cloud technology better, says Frank Frankovsky, Facebook’s director of technical operations and one of the founding members of the Open Compute Project. He’s working to encourage other companies to contribute improvements to Facebook’s designs.
Among the partners: chip makers Intel and AMD, which helped Facebook’s engineers tweak the design of the custom motherboards in its servers to get the best computing performance for the least electrical power use. Chinese Web giants Tencent and Baidu are also involved; after touring Facebook’s Oregon facility, Tencent’s engineers shared ideas about how to distribute power inside a data center more efficiently. Even Apple, which recently launched its iCloud service, is testing servers based on Facebook’s designs. Eventually the Open Compute Project could exist independently of the company that started it, as a shared resource for the industry.
Facebook’s project may be gaining traction because companies that manufacture servers, such as Hewlett-Packard and Dell, face a threat as business customers stop buying their own servers and instead turn to enormous third-party cloud operations like those offered by Amazon. “IT purchasing power is being consolidated into a smaller number of very large data centers,” Frankovsky says. “The product plans and road maps of suppliers haven’t been aligned with that.” Being able to study the designs of one of the biggest cloud operators around can help suppliers reshape their product lines for the cloud era.
However, not everyone wants servers to run just like Facebook’s, which are designed specifically for the demands of a giant online social network. That’s why Nebula, which offers a cloud computing platform derived from one originally developed at NASA, is tweaking Facebook’s designs and contributing them back to the Open Compute project. Nebula CEO Chris Kemp says this work will help companies that need greater memory and computing resources, such as biotech companies running simulations of drug mechanisms.
Larry Augustin, CEO of SugarCRM, which sells open-source cloud software to help businesses manage customer relations, sees challenges for Facebook’s project. “There have always been efforts on open hardware, but it is much harder to collaborate and share ideas than with open software,” he says. Nevertheless, Augustin expects the era of super-secret data center technology to eventually fade, simply because the secrecy is a distraction for businesses. “Many Internet companies today think that the way they run a data center is what differentiates them, but it is not,” he says. “Facebook has realized that opening up will drive down data centers’ costs so they can focus on their product, which is what really sets them apart.”
This new data poisoning tool lets artists fight back against generative AI
The tool, called Nightshade, messes up training data in ways that could cause serious damage to image-generating AI models.
Rogue superintelligence and merging with machines: Inside the mind of OpenAI’s chief scientist
An exclusive conversation with Ilya Sutskever on his fears for the future of AI and why they’ve made him change the focus of his life’s work.
The Biggest Questions: What is death?
New neuroscience is challenging our understanding of the dying process—bringing opportunities for the living.
Driving companywide efficiencies with AI
Advanced AI and ML capabilities revolutionize how administrative and operations tasks are done.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.