The Facebook guide to global infrastructure optimization (and domination)

Facebook is teaming up with CA Technologies to make its datacenters more efficient with the goal of further diminishing any chance of an outage on the News Feed.


While more than a billion people are over-sharing on the front end, Facebook itself hasn't shied away from sharing about what goes on underneath the world's largest social network either.

What keeps that social network going is energy -- in all meanings of the word. But it's about being energy efficient that might be the real key to Facebook's dominance.

The Menlo Park, Calif.-headquartered company frequently offers tidbits and deep dives about its search algorithms and infrastructure -- much of which is based on open source technologies.

Facebook also happens to be one of the major forces behind the Open Compute Summit, an annual conference in the heart of Silicon Valley that is gaining in traction and dedicated entirely to open source software and hardware development.

Facebook's vice president of infrastructure, Jay Parikh, shared some more during Tuesday morning's keynote in San Jose -- notably about the social network's recent methods for global infrastructure optimization.

Facebook's famous CEO and co-founder, Mark Zuckerberg, has spoken frequently in the past about a goal to establish a global social graph , an ambitious plan to basically connect all the dots in the world feasibly possible.

The global hardware and datacenter infrastructure, therefore, could be construed as the blueprint to making that happen.

Parikh explained that as Facebook's userbase grew exponentially, the priority for years was simply to keep the site up and running.

The lesson learned, he continued, was that all problems are magnified when running at scale and that the impact of broken infrastructures will compound quickly overtime.

But the Facebook infrastructure team realized last year that it was time to make some big changes, Parikh admitted, starting with rethinking structure at the datacenter level.

"There was one constraint: move fast," Parikh remarked.

Along with touting the architecture of the platform, Facebook routinely boasts about the designs of its own datacenters themselves . Parikh asserted that Facebook has done a lot of work to minimize energy consumption, lamenting at the same time that it can be especially difficult to make use of renewable energies in the United States, which is where most of Facebook's datacenters are located at the moment.

While it builds most of these facilities itself from the ground up, powering a platform that remarkably avoids significant downtimes to boot, Facebook can't do everything itself.

One of the ongoing partnerships involves software management provider CA Technologies. Facebook is using CA's data center infrastructure management (DCIM) software to improve power efficiency by connecting the dots between energy data points stemming from IT and hardware sources.

Why does this all matter? Sure, at one level it's about saving money, Parikh acknowledged. But it's no trifle matter considering Parikh cited Facebook has saved over $1.2 billion thanks to these energy efficient strategies.

At the same time, the idea is that a more efficient datacenter will eventually result in a more reliable experience for everyone, further preventing any possible outages and downtimes that would surely have Facebook users expressing rage over on another social network.