The Google Computer
I never really considered it before, but this post made me think. In it, Rich Skrenta argues that Google is building a giant distributed computer, consisting of 100,000 (so far) relatively cheap PC type boxes. I knew that Google was running Linux on their search cluster, because posts sometimes show up on Linux Kernel about this or that bug discovered by, or improvement needed by, Google. But, I assumed, like everyone did, that they were running a fairly standard application-oriented cluster on these boxes.
However, Rich makes perfect sense in his post. Google isn’t really running on Linux, they’re running on a virtual machine that uses Linux as it’s host. Linux merely isolates the Google OS (dubbed GooOS) from the hardware variations on they’re ever changing, ever expanding, ever improving cluster.
Like all Geeks with OS pretentions, I’ve thought about giant clustered computers, and how to build them, and what to do with them. I’ve thought about the Internet becoming one giant cluster, that everyone just taps into. I’ve thought about ubiquitous computing, in the form of small computing elements scattered everywhere: stuck in walls, on streetlamps, in cars, carried by people, embedded in people; bascially, anywere they can get power. These computing elements would only need power and a wireless connection to the Mesh, and would join the giant Computer that would be the Internet. I didn’t think it would happen this quickly, tho.
Google’s already made a surprisingly large step in that direction. Now I’m really curious as to the actual composition and layout and usage of the Google computer. Inquiring minds want to know.
