Google Blogoscoped


google scalability conference

Andrew Hitchcock [PersonRank 10]

Sunday, June 24, 2007
15 years ago2,853 views

Hi guys.

I'm at Google's scalability conference in Seattle. I just got out of the Marissa Meyer talk and she mentioned Caleb and his gadgets. Also, here are some interesting facts I learned:

Largest GFS filesystem: 5+ PB
Largest BigTable database: 3+ PB

Also, one guy briefly mentioned their machine translation work. For every doubling of input data, the quality of translation only goes up by .5% (compare to a human translator).

Philipp Lenssen [PersonRank 10]

15 years ago #

Interesting. For comparison...

Andrew Hitchcock [PersonRank 10]

15 years ago #

To put those numbers in perspective, if they were using Apple's Xserve RAID, they would need 35 racks to reach a petabyte. Since GFS stores three copies of the data, that would be 105 racks.

Also, there was a half-presentation by a guy that works on Google Talk. He mentioned how they stress tested their infrastructure before rolling out Talk in Gmail and Orkut. Before officially launching it, they put the actual service calls into their application, but didn't display anything to the user. So, for a few weeks before Chat in Gmail was launched, Gmail was actually logging you into the service behind the scenes. They would then slowly roll this out so they could scale up and test at full load before officially launching. I'll have to keep this in mind if I'm ever doing anything similar.


PS: I think those numbers are correct. You'll be able to verify when the videos are posted to Google Video.

Forum home


Blog  |  Forum     more >> Archive | Feed | Google's blogs | About


This site unofficially covers Google™ and more with some rights reserved. Join our forum!