Forget about online advertising for a few minutes. Try (if you can) to put aside questions about Google's (GOOG) sky-high stock price, the rumors about the company's foray into mobile telephony, or its plans to extend its influence on media.
Instead, think about Google as a star-studded collection of computer scientists who have access to a fabulous machine, a distributed network of data centers that behave as one. These globe-spanning networks of computers are known as "clouds." They represent a new species of global supercomputer, one that specializes in burrowing through mountains of random, unstructured data at lightning speed. Scientists are hungry for this kind of computing. Data-deluged businesses need it. What will Google do with its machine?
BusinessWeek writers Stephen Baker and Rob Hof sat down recently at Google headquarters with Chief Executive Eric Schmidt to talk about Google's machine and its venture with IBM (IBM) to extend Google-style cloud computing to the entire world. The project begins with a pilot program in six universities.
On cloud computing:
What [cloud computing] has come to mean now is a synonym for the return of the mainframe. It used to be that mainframes had all of the data. You had these relatively dumb terminals. In the PC period, the PC took over a lot of that functionality, which is great. We now have the return of the mainframe, and the mainframe is a set of computers. You never visit them, you never see them. But they're out there. They're in a cloud somewhere. They're in the sky, and they're always around. That's roughly the metaphor.
On Google's place in cloud computing:
Google is a cloud computing server, and in fact we are spending billions of dollars—this is public information—to build data centers, which are in one sense a return to the mainframe. In another sense, they're one large supercomputer. And in another sense, they are the cloud itself.
So Google aspires to be a large portion of the cloud, or a cloud that you would interact with every day. Why would Google want to do that? Well, because we're particularly good at high-speed data and data computation.
On Google's software edge:
Google is so fast because more than one computer is working on your query. It farms out your question, if you will, to on the order of 25 computers. It says, "You guys look over here for some answers, you guys look over here for some answers." And then the answers come back very quickly. It then organizes it to a single answer. You can't tell which computer gave you the answer.
On the size of cloud computing:
There's no limit. The reason Google is investing so much in very-high-speed data is because we see this explosion, essentially digital data multimedia explosion, as infinitely larger than people are talking about today. Everything can be measured, sensed, tracked in real time.
On applications that run on a cloud:
Let's look at Google Earth. You can think of the cloud and the servers that provide Google Earth as a platform for applications. The term we use is location-based services. Here's a simple example. Everyone here has cell phones with GPS and a camera. Imagine if all of a sudden there were a mobile phone which took picture after picture after picture, and posted it to Google Earth about what's going on in the world. Now is that interesting, or will it produce enormous amounts of noise? My guess is that it'll be a lot of noise.
So then we'll have to design algorithms that will sort through to find the things that are interesting or special, which is yet another need for cloud computing. One of the problems is you have these large collections coming in, and they have relatively high noise to value.
In our world, it's a search problem.
On Google becoming a giant of computing:
This is our goal. We're doing it because the applications actually need these services. A typical example is that you're a Gmail user. Most people's attachments are megabytes long, because they're attaching everything plus the kitchen sink, and they're using Gmail for transporting random bags of bits. That's the problem of scale. But from a Google perspective, it provides significant barriers to entry against our competitors, except for the very well-funded ones.
I like to think of [the data centers] as cyclotrons. There are only a few cyclotrons in physics and every one of them is important, because if you're a top flight physicist you need to be at the lab where that cyclotron is being run because that's where history's going to be made, that's where the inventions are going to come from. So my idea is that if you think of these as supercomputers that happen to be assembled from smaller computers, we have the most attractive supercomputers, from a science perspective, for people to come work on.
On the Google-IBM education project:
Universities were having trouble participating in this phenomenon [cloud computing] because they couldn't afford the billions of dollars it takes to build these enormous facilities. So [Christophe Bisciglia] figured out a way to get a smaller version of what we're doing into the curriculum, which is clearly positive from our perspective, because it gets the concepts out. But it also whets the appetite for people to say, "Hey, I want 10,000 computers," as opposed to 100.
On meeting young engineers who talk to him about their 20% time projects (where Google employees are given time to work on their own projects):
You basically never know when you're going to meet the next Christophe. They might show up in the parking lot. They might show up on the walk. Here's an example. Somebody walks up to me and starts showing me this demo. It was in AdSense, and all of a sudden I realize that he had invented a billion-dollar business. I'd rather not go into the specifics. And I said, "How long have you been working on this?" He was like, "Oh, you know, about a month." I said, "Is this your 20% time? Have you told anyone about it?" He said, "Yeah, I was going to tell my manager, but I was afraid he'd tell me to stop." And I said, "Let me talk to the manager." This project really has potential for some really significant applications in the advertising world.
On other favorite 20% projects:
My favorite one is spelling correction. It's a bizarre story. The fellow, an undergraduate at Berkeley, wasn't quite sure what he wanted to do. He interviewed O.K., was obviously brilliant, kind of unfocused. He comes in, hangs out with his friends, and says, "Why don't we apply artificial intelligence technology to spell correction?" So he invents, by himself, the spelling corrector we use today. The algorithm that he invented has been explained to me, by him, by others, a couple of times, and I still don't really understand it. It's magic. Magic is science insufficiently explained.