Silicon Valley IT giants are typically mum about saying much about their massive data centers, but Google is on a media campaign right now, letting reporters in, including Wired’s Steven Levy. Google indexes 20 billion web pages a day, handles 3 billion search queries a day, and streams millions of Youtube videos every day. It’s data centers are its backbone, the core of its operations.
One of the interesting bits in the Wired story was that as early as 1999, Google’s Urs Holzle was tinkering around with building his own server setups. It’s long been rumored that the big players like Google, Facebook and Amazon want to go around the OEMs like Dell and HP, and build their own servers. And the story of why Holzle originally did it, to take unnecessary parts off of the motherboard like graphics processors, makes a lot of sense. The other big disclosure is that the company has mulled over building its own chips. There are some big R&D and patent challenges to Google doing this, but it’s not impossible and even more reason that Intel and AMD are going to allow its major customers to dictate more and more processor design.
The article closes on a competitive note, despite Google’s new willingness to let a reporter into its facilities. When asked what’s in Google’s future data center, Holzle said, “This is one thing I can’t talk about, because we’ve spent our own blood, sweat, and tears. I want others to spend their own blood, sweat, and tears making the same discoveries.” And that’s the biggest reality for webscale IT companies that people tend to think of a consumer web companies. Their real competitive advantages are actually based in infrastructure, doing compute tasks efficiently and at low cost. Because every search query that runs on Google’s servers costs the company money.