How is the Open Compute Project affecting today’s data centers
FYI, this story is more than a year old
Recently we ran an Open Compute Project (OCP) Engineering Workshop during a DatacenterDynamics event in London, which I attended in my role as Joint Chair of the OCP data center design project.
I was asked how I thought OCP was affecting today’s data centers.
If you’re not already aware, OCP was founded by Facebook in 2011. Today it is a growing community of like-minded engineers engaged in a mission to design and enable the delivery of the most efficient server, storage and data center hardware designs available for scalable and cloud computing.
I believe one of the main impacts OCP is having is to question the way we do things. Instead of taking a traditional siloed approach where, e.g., facilities build the data center, IT adds the servers and storage, OCP evaluates the influence on efficiency of all the components within the data center ecosystem.
This approach has already demonstrated how it can lead to optimized energy consumption as well as material use.
Take for example IT hardware. By designing out 6lbs of excess materials per server, OCP has not only reduced the cost of production, assembly and shipping, but also the eventual impact of disassembly and disposal or recycling.
The elimination of unnecessary panels around the server chassis means more efficient cooling and allows the servers to operate in higher temperatures.
The Open Data Center facility by Facebook has been designed in tandem with Open Compute servers to maintain this focus on reducing environmental impact. The data center maximizes mechanical performance as well as thermal and electrical efficiency.
In the US, it accepts 277 volts AC current for improved PUE as more energy makes it from the utility via the data center to server components.
You can see how integrated thinking works out at Facebook’s Prineville, Oregon, data center. Here both the data center and server designs have been integrated and built in tandem, leveraging the strengths of both the physical infrastructure and the hardware it supports. As a result of the OCP, Prineville is one of the most energy efficient data centers in the world with a PUE of 1.09.
When the OCP movement started, it was mainly the web giants that were able to build their own servers according to specific applications. However, as the development of standards based designs goes forward, the risks associated with self-build becomes mitigated across the community.
We noted that Fidelity, the financial services organization has deployed OCP hardware in their data center. I can imagine that smaller organizations will want to access the cost savings and efficiencies of OCP.
Although we are right at the start of the market, there are very high levels of interest in OCP and ideas like the use of Lithium Ion technology for power back-up have certainly stimulated a raft of announcements among data center physical infrastructure manufacturers.
Challenging legacy approaches is important if we’re going to move the industry forward and I believe that the collaborative, community approach where anyone can contribute is going to have a prominent place.
Article by Robert Bunger, Schneider Electric Data Center Blog