The Open Compute Project, the Facebook-helmed community initiative with the self-appointed mission of promoting more efficient, environmentally-friendly data center designs and architectures, used this week’s Open Compute Summit in San Antonio to launch the Open Rack specification – a standard for server rack design that promotes scalability in a way that the legacy approach can’t.
The “open” part of Open Rack comes from the standardization of server height, depth, mounting and cabling schemes, and connectors, as well as width, which is the only thing that the EIA 310-D specification calls for. See, as the Open Compute Project explains in a blog entry, EIA 310-D came to be in the 1950′s, before “scalable computing” was a concept, let alone a reality. The result was a data center equipment market where vendors were free to create their own standards, often locking customers in to one vendor or another’s data center designs.
Enter Open Rack, which essentially wants to do for modern data center hardware what OpenStack wants to do for cloud platforms. A new standard means the chance to solve not only vendor lock-in, but also architect better solutions that are more suited for meeting the cloud-oriented challenges of the modern IT landscape. It’s the Open Compute Program’s most extreme initiative to date.
In short, the Open Rack standard is slightly taller, for better airflow management and cabling both, the equipment space is 21″, but can be made to fit existing 19″-wide hardware, while maintaining the current standard 24″ column width. The Open Compute Project is boasting that the new standard makes 87.5% efficient use of space, up from the current standard, which only offers 73%.
Another key selling point for Open Rack is how it addresses TCO: Each compute component is designed to replaced according to its own lifecycle, so rather than swap out a server every 2.5 years like clockwork, it’s possible to only upgrade or replace as needed and squeeze performance out of a purchase for as long as ten years, in some cases. Oh, and it doesn’t use cables for power: bus bars at the back of the rack replace in-built power supplies, just plug it in and you’re good to go.
Obviously, there’s a little more to it, but that’s the gist of how the Open Compute Project is looking to revolutionize the data center equipment standard. And with heavyweights like HP, AMD, Fidelity, Salesforce.com, VMware, Canonical, AMD, Sun Microsystems, Intel, Avnet, Alibaba, Supermicro, Cloudscaling, Quanta and obviously Facebook on board and building solutions to run on OCP-compliant hardware, there’s some market muscle there.