Open Compute Summit: What The Future May Hold

Printer-friendly version Email this CRN article

The Open Compute Project has generated lots of interest in developing open power efficiency standards for data centers since Facebook launched it in June, 2011. The group heated up this week with news involving AMD, Fusion-io, Intel, Rackspace and others in attendance at the Open Compute Summit, its semi-annual confab.

Topping the list were AMD and Intel, both of which revealed plans last year to develop motherboards that comply with OCP's Open Rack specification -- AMD with its Roadrunner motherboard design, and Intel with Decathlete. Both companies showed working products this week.

Another big get for OCP was Rackspace, which announced this week that it would build an all new infrastructure based on OCP specifications. Other converts include Riot Games, which revealed that it would purchase OCP-based systems from Hyve Solutions, a contributor to Facebook's vaunted Prineville Data Center.

[Related: Facebook Launches New Social Search Engine]

The organization also announced the addition of more than a dozen new members, including Applied Micro, AMD, ARM, Calxeda, EMC, Fusion-io, Hitachi, NTT Data, Orange and SanDisk.

Fusion-io also contributed the design for the 3.2TB ioScale card. Other new OCP-compliant "open source hardware" designs include a storage box that Hyve built for Open Rack and Rackspace's infrastructure plans. Facebook itself designed specs for an all-flash database server, code-named Dragonstone, updated its Web server, code-named Winterfell, and provided modifications to Open Rack and Open Vault specs for cold-storage environments.

Longer-term, the Open Compute Platform is developing a means to decouple the parts of a motherboard such as its processor, networking expansion busses. "Much of the hardware we build and consume is highly monolithic," read a Jan. 16 news release related to the summit. "This leads to poorly configured systems that can't keep up with rapidly evolving software and waste lots of energy and material."

The solution, according to the OCP, is to "disaggregate some of the components of these technologies from each other" so as to allow systems to be built for their intended workloads with components that can be replaced or updated independently of each other. Stepping up to this challenge, Intel said it plans to contribute a 100-Gbps interconnect technology, and Facebook will develop a common-slot architecture that it nicknamed "Group Hug" with buy-in from AMD, Applied Micro, Calxeda and Intel.


Printer-friendly version Email this CRN article