Google garage

by Shane Schick:

The search giant builds rather than buys, but we may end up footing the bill

He was a smart guy, a good writer and a valued colleague, but for some reason he carried a wallet made out of duct tape. It consisted of little more than a few pieces of the material stuck together, but he had designed it himself and it worked. We laughed at how proud of it he used to be, but I wouldn’t laugh today. I’d suggest he go work for Google.

Any corporate enterprise that’s struggling with a build versus buy decision regarding their IT infrastructure would have been interested to read a feature story in The New York Times recently which unveiled the weirdly inventive way Google keeps its search engine running. It’s a fairly straightforward policy: whatever the company needs, it makes. This includes everything from custom-made servers and racks of computers held together with Velcro (a step up from duct tape, but still) to its own Alpha-influenced microchips. Lots of companies start about their origins in a basement or garage, but from the sounds of it Google has never really left.

Perhaps Google wants to eschew the traditional interdependency that marks the relationships between hardware and software firms. Microsoft and HP may compete on some system management software products, for instance, but Microsoft is probably a loyal customer of HP server and storage products, just as a lot of HP employees probably use Windows. Google’s partnership with Sun, on the other hand, didn’t seem to hinge on its use of Sparc workstations per se.

Even though it flies in the face of our usual notions of core competency, Google’s expansion into more areas of information management – e-mail and spreadsheets, for instance – may require it to create something beyond the state of the art. As more users take advantage of its services, they will depend on its uptime in ways no search engine has before. But the dependency may not stop there.

Although there is something charming about the elf-like work ethic behind Google’s approach, it begs the question of whether other organizations should follow its example. The company claims it saves money and meets its unique needs by cobbling together everything except its own desks and office chairs, but it’s hard to believe it couldn’t achieve the same results with other vendor’s equipment. Banks, insurance companies and other firms may write their own software code, and some may configure their hardware in relatively original ways. They aren’t, however, necessarily seeking patents on their designs, which Google is. That hints at an agenda to collect and capture its proprietary ways of architecting a data centre, rather than sharing them with the rest of the industry.

Only a few companies in the world could afford to do the kind of in-house engineering that Google is doing, although the results of its experimentation could create benefits for IT managers and CIOs everywhere. Rather than contributing to or supporting industry standards around the data centre, though, Google’s do-it-yourself dogma appears aimed at circumventing them. One day we may be looking to Google in search of answers much different than the basic ones we do today, and the results may cost us a lot more.