We just caught up with Urs Hölzle, who was one of Google’s first 10 employees and today is a senior vice president in charge of Google’s vast technical infrastructure—the data centres and networks on which Google’s services run.
Today at the I/O conference in San Francisco, Hölzle unveiled Google Compute Engine, a service which gives third-party developers much more direct access to that infrastructure.
We asked him how Compute Engine stacked up against competitive offerings, like Amazon’s EC2.
He told us he thought the real opportunity was what he estimated as the 99% of computing power that doesn’t run on the cloud today—for example, in proprietary supercomputing clusters, a big potential target for the high-performance computation service he showed off today.
We also asked if he thought Google’s infrastructure could evolve to serve primarily outside customers. Today, Google infrastructure overwhelmingly runs its own services. But Amazon, because of the growth of Amazon Web Services, now devotes a majority of its computing resources to external customers.
Could Google get there? Hölzle cited a recent Stanford University study which reported that Google has about 1 per cent of the world’s servers.
“If we can do 1 per cent, can we get to 5 per cent? Easily,” Hölzle said.
We’re inclined to believe him: Hölzle is the one who’s built Google’s infrastructure going back to the company’s earliest days.