How does google work?
 

I noticed that Google.com claims to search about 8 billion pages.

How is it that when I click Search it takes less than a second to search all 8 billion pages?

Even if an entire page could be searched in one clock cycle(which it can't be), the fastest servers I have seen are 3 gigagherts. At one page per cycle that google.com would need 8 gigahertz.

Obviously it takes a lot more than 1 cycle to search a page, a lot more than 1. So how does google.com (or any other search engine) come up with such incredible power?

Does it have a crazylinked database where everyword in the database has like a million links?

Curiosity has gotten the better of me.

Thanks,

David

PIGEONS!!!! check out this link!!!!
http://www.google.com/technology/pigeonrank.html

Moved: This thread is more suitable in General and has been moved accordingly.

Not sure if it's what you want, but I have a .pdf of the Google Cluster Architecture.

http://www.digitalcarnage.net/files/...chitecture.pdf

Google Uses the power of clusters. Basically, they have hundreds, possibly thouthands of top-end servers with multiple processors linked together, essentially combining their processors into one. So, for an example, if they were to only have say 200 computers clustered together, each with 2 2Ghz processors that would be 400 2ghz processors, or 800Ghz of processing power. In reality, google likely has far more servers with more, faster processors per server. Anyway, that is just a basic explanation of how google searches so fast. Hopefully another memmber will have another link to a better explanation.

Google has one of the largest Linux farms anywhere. I think they have a custimized version of redhat . with that much processing power
they can do pretty much anything they want.


http://www.internetweek.com/lead/lead060100.htm


-Nex6

Google's cluster

* 719 racks
* 63,272 machines
* 126,544 CPUs
* 253,088 GHz of processing power
* 126,544 GB of RAM
* 5,062 TB of hard drive space

November 2004: 8,058,044,651 web pages, 880,000,000 images, 845,000,000 messages, 4,500 news sources

Holy Crap
 

I am glad I am not paying that electricity bill....


Thanks guys! I knew that Google had some crazy clusters. But WOW! that is a lot of machines.

David

Wow,*very cool info*, where did you find it?

mark

I've seen those figures around the net and AFAIK they are just estimates.

I also presume that what those clusters are running is an searchable index of keywords in a generated database. Not actually searching each and every page.

http://en.wikipedia.org/wiki/Google

-Joey

I like tacos.

Again, to clarify, if you read the info in that link, it says that those statistics are only an estimate or googles cluster size. Th only people who would know 100% would be Google employees.

Re: How does google work?
 

afaik (and this is both rough and from memory)
Thankfully, that's not how it works. Pages don't get searched when you search, instead when pages are added, hashes of strings on the page are stored, and that's what is searched. Searching actual pages would be far too slow (impossibly slow), although it would mean wildcard searches would possible, something google (and other search engines) can't do.