Inside the Googleplex

Web computing could make a CIO's life much, much easier. Applications will be hosted by third parties that can support an almost unlimited number of users and store almost unlimited amounts of data. CIOs will no longer have to maintain large databases or expensive desktop software. And because these applications live on the Web, your employees will be able to access them with any device that supports a browser.

In this environment, upgrades are easy. Instead of an IT department pushing out a patch to every user, or installing software from a CD onto machines one by one, the application provider updates it once, and everyone gets the new version the next time they log on. That's what Google did when it integrated chat into Gmail. CIOs will no longer have to buy large, packaged applications, like an ERP system, with hundreds of functions they'll never use. Online applications will allow them to pick and choose the services they need. And because there are no legacy investment costs in each service, switching between services — and providers — will be easy. You just point your employees to another Web site.

Sounds great. So why isn't everyone doing it?

"Building a suitable Web computer is really hard to do right now," says Paul Kedrosky, executive director of the William J von Liebig Centre, a venture capital firm that commercializes technologies developed at the University of California, San Diego. "But it was also really hard to create a single-circuit board. Then it turned into a commodity." Like the circuit board, most experts believe Web computing will also turn into a commodity.

But today, Google is the only company that's got it (for companies that are close, see "Google's Progeny", below). And the reason for that is under the hood.

Google started out with two advantages. First, it was a technology greenfield, unencumbered by a legacy architecture. Second, its founders were cash-poor grad students who needed to make their system run with tools they could afford. Google co-founders Sergey Brin and Larry Page taught themselves how to turn cheap, off-the-shelf PCs (which, legend has it, they liberated from the Stanford computer science department's loading dock) into powerful servers. And they connected these servers to one another with Linux, a free operating system that they customized over time to meet their needs.

By 2002, Google's homemade, commodity-based servers outperformed comparable equipment from IBM that cost $US18,000 for 360 gigabytes of storage; Google spent about $US1000 for the same amount.

Google treats its infrastructure as a closely guarded secret. It doesn't allow outsiders into its data centres, so any description of the company's infrastructure is an educated guess. Philipp Lenssen, who runs the Google Blogoscoped Web site, says that the last semi-official count of the company's servers was 100,000. Stephen Arnold, an independent consultant and author of the e-book The Google Legacy, suspects that it is now more than 150,000, spread across 24 data centres — making Google the third largest server manufacturer in the world, according to Arnold. The power per dollar that Google is able to achieve compared with its competitors explains the speed of its Web searches, the 2.5 gigabyte inboxes it gives each Gmail user, and other storage-intensive initiatives, such as Google video, which essentially invites anyone to upload a video to Google's servers.

"Our infrastructure is an important competitive advantage for us," allows Dave Girouard, general manager of Google's enterprise division. "It has incredible reliability and scalability, and the cost per user is incredibly low."

Behind the Fun and Games

Given the combination of Google's secrecy and its relatively anodyne media image — which the company works hard to foster with its "Do No Evil" corporate motto — it's no surprise that people looking at the company just see a search engine, some snappy consumer applications and no coherent commercial strategy. "But what you see on the surface is just what they choose to show you," says Lenssen. "It's like they are holding a magnet beneath a table. All you see is a metal ball moving around on top. Because you can't see the magnet, you assume the ball is moving randomly."

Less colourfully put, because Google's strength lies not in the applications it shows the world but in the architecture it doesn't, the company's business strategy is not readily apparent. But according to Lenssen, people who look at Google and see nothing more than random growth are making a big mistake.

Most of the high-tech industry is just starting to realize this. "IBM executives in the early 1980s didn't understand what Microsoft was," says Arnold. "Now Microsoft is in the same spot, and they are trying to understand what Google is. And they're having a hard time."

Arnold, who has spent the past year performing competitive analyses of Google for large technology companies, says: "People don't want to believe things that run counter to their vested interest." Meaning, in this case, a way of computing that is not dependent on the client/server architecture. "They refuse to recognize that a search engine with ads has got something different enough to alter the rules," he says.

But it does.

Google has slipped under the radar as a potential enterprise computing powerhouse because it hasn't tried to compete directly with Microsoft or any of the other big technology vendors. Google could offer an unlimited number of applications on top of its infrastructure. The ones it has chosen to unveil — search, e-mail, maps, blogging tools, chat — aren't exactly Office. But Google has all the pieces. Gmail contains many of the rich text features, such as font selection, highlighting and alignment, found in Word; the Google search bar can perform the same mathematical functions as Excel; and Picasa, Google's online photo-editing service, has some of the presentation capability of PowerPoint.