The Disruptor: Frank Frankovsky, VP Infrastructure for Facebook
by Stacey Higginbotham
Remember when the idea of personal computers finding a home in the data center was crazy? So does Frank Frankovsky, a VP of infrastructure for Facebook, who back in 1990 was working at Compaq helping build out the first servers based on Intel chips that were previously reserved for home computers. Today Frank Frankovsky is set to implement another data center revolution: dismantling the humble server he helped create.
Frankovsky, who was hired by Facebook in 2009, helped build the servers that became the model for the Open Compute Project. His role since has expanded to encompass not just the servers, but the hardware design and supply chain for Facebook’s physical infrastructure, a daunting task considering the site serves 901 million users, has more than 100 petabytes of user photos and videos, and supports millions of lines of code.
But for Frankovsky — who was originally a business major who just happened into the computer trade thanks to an internship at Compaq during college — is up for the challenge. In fact, he’s rethinking the challenge in terms that could take the data center and turn it into a giant, massively modular computer, designed solely to run Facebook’s software.
Think of it as the ultimate integrated system and a perfect case study for the emergence of software defined infrastructure. In the past few years Facebook has been working to design and build the best, most energy efficient system to run its hardware, integrating both the server infrastructure to the physical data center.
The man behind all this novel server design was Frankovsky. In April 2011 the social network announced the results of that project and created the Open Compute Project to push the idea (and the hardware) beyond Facebook. And in this way Frankovsky is influencing an evolution of infrastructure beyond the top social media site, and he’s taking his dream of a massively engineered data-center-sized computer to banks, webscale companies and cloud providers.
Where the previous focus for Open Compute and Frankovsky’s efforts were to build an energy efficient data center that optimized every screw and component, he’s now turning to optimizing everything in the newly designed data center to function efficiently. We’re moving from building a data center at scale to operating a data center efficiently at scale, said Frankovsky.
That level of efficiency means finding modular parts that can be replaced at the CPU-level, the networking level and at every other level. IT will also encompass energy monitoring and scheduling software that helps schedule jobs when power costs are lower, if possible, and will also help ensure that computers aren’t left to idle.
These are hard problems to solve, which makes redesigning the hardware look like low-hanging fruit. But with the demand for computing and infrastructure at an all time high, and only expected to rise, Frankovsky knows what he has to do and plans to make it happen. For Facebook, where the cost of computing directly affects the cost of delivering ads that make it money, the infrastructure is just as important as something like just-in-time inventory management is to a company like Dell.
And as more and more companies forgo owning the bulk of their IT infrastructure, Frankovsky stands to help influence how the majority of the world’s future compute providers build out their infrastructure. He’s already changed the relationship between hardware suppliers and their largest customers and is looking to shake things up even further by changing how networking, job scheduling and even motherboards work.
He may have helped build the current status quo during his long ago days at Compaq, but now he’s trying to tear that apart.