Microsoft, which in January contributed the server design it uses to run its massive cloud and hyperscale data center environments to the Open Compute Project, is now releasing the second generation of the server specification.
Officials with the software giant made the announcement Oct. 30 during a keynote address at the Open Compute Project (OCP) European Summit in Paris, where the company also is showing off a range of designs and platforms based on the second-generation Open CloudServer (OCS v2) specification.
Among the key innovations in this latest generation spec is a dual-processor design leveraging Intel’s Xeon E5-2600 v3 “Grantley” chip—which will enable 28 processing cores per blade system—use of 40 Gigabit Ethernet networking technology and support for flash memory, which offers high bandwidth and lower costs than traditional hard drives.
“Overall, the OCS v2 specification is the convergence of a number of design points,” Kushagra Vaid, general manager of server engineering at Microsoft, said in a post on the company blog. “Meeting these design points in one server is critical for Microsoft to meet its service delivery and datacenter operations goals, and also the needs of the community who’ll use OCS v2 for their own datacenter operations.”
Large Web companies like Microsoft, Google, Facebook and Amazon run massive data centers with huge numbers of servers, and to increase performance and density while saving money on power costs and floor space, they have turned to building their own infrastructures—including servers, storage appliances and networking gear—rather than relying solely on off-the-shelf gear from the likes of Dell, Hewlett-Packard and Cisco Systems.
Facebook in 2011 drove the launch of the OCP, an effort designed to apply the same open development practices used in software to drive the design of highly cost-effective and energy-efficient hardware in the data center.
Microsoft, which runs more than 1 million servers in its data centers, joined the consortium in January when it submitted the first Open CloudServer spec. Microsoft not only runs a huge and growing cloud operation, but also a large traditional data center environment.
According to Vaid, the software vendor has invested more than $15 billion in its worldwide cloud infrastructure, which offers almost 200 cloud services to 1 billion customers and 20 million businesses in more than 90 markets. Given that, a key point of interest was ensuring the server design could run an array of cloud services.
Over the last six months, Microsoft has tested the OCS v2 design in its data centers “from powering IaaS [infrastructure-as-a-service] and PaaS [platform-as-a-service] services in Windows Azure, to hosting e-mail and collaboration services in Office 365, to hosting latency-sensitive gaming services in Xbox Live. Converging onto a unified, flexible design allows us to optimize the economics of our supply chain, while delivering a diverse array of cloud services from one underlying server platform.”
The other design points where to make sure the systems could be used anywhere in the world—they can meet a range of electrical standards, safety regulations and regulatory requirements around such aspects as rack size and cooling parameters—and that they offer the best performance while keeping down total cost of ownership.
Along with the dual-processor design, 40 GbE capabilities and support for flash memory, the OCS v2-based systems offer flexibility in the core design to enable the use of an array of technologies—such as field-programmable gate array (FPGA) accelerators to make it easier for users to optimize the systems for their needs—compact power supplies that deliver up to 1,600 watts, and support for high-memory configurations of 128GB, 192GB and 256GB.
At the OCP’s European summit, Microsoft demonstrated OCS v2 designs from such partners as Quanta, Wiwynn and ZT Systems, according to Vaid.