Google datacenters

HHOSUG

Quicksearch

Navigation

Events

< A380 MSN004 over Berlin Tempelhof | links for 2008-06-02 >

Monday, June 2. 2008

Interesting article about the failure rates at the datacenter at Google:

In each cluster's first year, it's typical that 1,000 individual machine failures will occur; thousands of hard drive failures will occur; one power distribution unit will fail, bringing down 500 to 1,000 machines for about 6 hours; 20 racks will fail, each time causing 40 to 80 machines to vanish from the network; 5 racks will "go wonky," with half their network packets missing in action; and the cluster will have to be rewired once, affecting 5 percent of the machines at any given moment over a 2-day span, Dean said. And there's about a 50 percent chance that the cluster will overheat, taking down most of the servers in less than 5 minutes and taking 1 to 2 days to recover.

Posted by Joerg Moellenkamp in English, The IT Business at 09:59 | Comments (2) | Trackback (1)
View as PDF: This entry | This month | Full blog

Trackbacks

Trackback specific URI for this entry

PingBack

Weblog: o9y.net
Tracked: Jun 30, 18:51

Comments

Display comments as (Linear | Threaded)

I have a hard time being very excited about Google's data centers. It's interesting, but the problems they are solving are, well, completely different than those that need to be solved for most businesses.

For Google, if they lost 5% of their data it doesn't matter as search results don't have a "right answer." But a company running welfare claims losing 5% of their data means that 5% of the recipients of that welfare don't get to eat for a week. It's a totally different game.

How much do you want to bet that Google's accounts receivable, payroll, and HR systems run on COTS ERP software and typical robust server hardware deployed in a way that would surprise no one?

#1 Anonymous on 2008-06-02 19:16 (Reply)

As far as i know they use very reliable hardware for that task .. i remember a job offering for a Sun Admin

Of course it´s not the way for all tasks, but i think at least for some tasks this solution is feasible for enterprise work. The basic idea behind this concept at google is that they don´t loose data and mostly the can still fulfil the request ...

#1.1 Joerg+M. (Homepage) on 2008-06-02 21:15 (Reply)

Add Comment

Name
Email
Homepage
In reply to
Comment	E-Mail addresses will not be displayed and will only be used for E-Mail notifications To prevent automated Bots from commentspamming, please enter the string you see in the image below in the appropriate input box. Your comment will only be submitted if the strings match. Please ensure that your browser supports and accepts cookies, or your comment cannot be verified correctly. Enter the string from the spam-prevention image above: Standard emoticons like :-) and ;-) are converted to images. Enclosing asterisks marks text as bold (word), underscore are made via _word_.
	Remember Information? Subscribe to this entry

View as PDF: This entry | This month | Full blog

Competition entry by David Cummins powered by Serendipity v1.0

Links in this article

the failure rates at the
datacenter at Google

The LKSF book

The book with the consolidated Less known Solaris Tutorials is available for download here

Twitterfeeds

twitter.com/c0t0d0s0
just blogged: links for 2009-07-04 http://tinyurl.com/mrgt3t

twitter.com/codenews
6853745 Same ereport is generated every 10 seconds automatically in looping after error injection http://tinyurl.com/nwldlg

twitter.com/SunPatches
122186-13 - McData fibre channel director switch firmware e/osn 9.9.3 for i10K. Available since Jul/02/09. http://tinyurl.com/nmu5zs

twitter.com/SolPatchesX86
121431-38 - SunOS 5.8_x86 5.9_x86 5.10_x86: Live Upgrade Patch. Available since Jul/02/09. http://tinyurl.com/m44dul

twitter.com/SolPatchesSPARC
121430-37 - SunOS 5.8 5.9 5.10: Live Upgrade Patch. Available since Jul/02/09. http://tinyurl.com/mr4yqe