published @ [searchlores.org] in June 2000
[Back to main s.e.]

Search engine Google has deployed 4,000 Linux servers, with plans to increase to 6,000 this year, making it possibly the largest Linux installation in the world.

Google said it turned to Red Hat Linux primarily because of the cost. The OS itself costs nothing, compared with $500 to $900 per server for Windows servers.

And the hardware is also cheap. Red Hat runs on commodity white-box PCs rather than more expensive RISC Unix servers.

"The hypertext analysis is computationally expensive," said Sergey Brin, founder and president of Google.com. "We need to have an efficient system for doing that. That's why we use a lot of cheap PCs. It's a cheaper platform. The dollar per MIPS is better for PCs."

Google ranks the importance of a target Web page by determining how many pages link to that page, and how important the Web pages are which link to the target page. It also uses hypertext analysis to determine where on a Web page a keyword is located.

That's extremely computationally intensive, with 500 million variables and 2 million terms in a search equation to index the Web, performed about every month, resulting in about 1 TB of data to index 300 million Web pages.

Support was another factor in choosing Linux, Google said. The company has Linux expertise in-house, and values the ability to look at the source code to correct problems, rather than having to rely on a vendor. And where the in-house expertise fails, Google has found the Linux community responsive with fixes.