Certainly more than one, but not necessarily "lots". The Google File System (GFS), not to be confused with Global File System, was written by Google and is a fully distributed filesystem. Each 64MB data block is replicated to 3 other servers, with each server being connected to physically separate layer 2 switches and located in different racks. The system is completely self-healing. If you loose a data block, disk or entire server it automatically rebuilds the data on another server. The entire server is a Field Replaceable Unit. Simply plug the server into the cluster and it is automatically provisioned, joins the cluster and starts serving queries. Each cluster has up to 2000 servers which are simply vanilla x86 hardware with IDE hard drives. Each nodes workload is specifically balanced to maintain a 2Mbps stream allowing large clusters to sustain 2 Gbps I/O rate.
BTW, I googled for all this info a few days ago when the thread started. Took a few minutes of refining the query to weed out the cruft, but high level information is out there. -Sam -----Original Message----- From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of John Summerfied Sent: Thursday, May 11, 2006 10:16 PM To: [email protected] Subject: Re: Fw: [LINUX-390] Google out of capacity? Yu Safin wrote: >> > It seems we covered the CPU side pretty well, but what about the DISK > side. are they using SAN? > I think someone mentioned earlier on that the disk is in each box but > later someone else mention that it is not but external. > wouldn't a lot of the same arguments apply to the storage? > it seems like the data is not that important but do they run any kind > of backups? I imagine they have lots of versions of "the database," and that the loss of one isn't very important. This would go to explain reports by others that search results vary depending on where in the world are when you search. -- Cheers John -- spambait [EMAIL PROTECTED] [EMAIL PROTECTED] Tourist pics http://portgeographe.environmentaldisasters.cds.merseine.nu/ do not reply off-list ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
