The phrase “open source” nearly always modifies “software,” but it doesn’t have to be that way. Cloud-related hardware, too, can be developed using the open-source model, and Microsoft says it has a better way to do that.
In a blog post today, Azure hardware-infrastructure general manager Kushagra Vaid laid out the basics of Project Olympus, an effort to introduce open-source hardware designs far earlier in the design process than is conventional.
Since 2014, Microsoft has been a member of the Open Compute Project — an industry group including Apple, Dell, Facebook, Google, Intel, Juniper Networks and Nokia — that shares designs for data-center hardware. The OCP norm is to share designs when they are production ready, Vaid said. At that point, it’s too late for major changes. So Microsoft will begin contributing designs when they’re about halfway complete.
That will let the community download, modify and even fork (vary from) the design, just as happens with open-source software, Vaid said. The change could lead to faster delivery of new hardware, he said. Bill Carter, CTO for the Open Compute Project Foundation, in a release called the change “unprecedented in the history of OCP.”
Under Project Olympus, Microsoft plans to contribute a new motherboard, a high-availability power supply with included batteries, a server chassis, high-density storage expansion, a new rack power distribution unit and a standards-compliant rack-management card. These modular building blocks can be used independently, Vaid said.
More than 90 percent of the servers Microsoft now buys are based on OCP-contributed specs, Vaid said. He said he plans to share more information on Project Olympus at Datacenter Dynamics: Zettastructure, a conference convening Tuesday in London.