obviously CBS buying last.fm for £140m didnt leave enough for the power supply budget...
They have a "datacenter"?!
I thought everything ran on one creaky Windows 98 PC. That was the response level anyway.
You've never seen lastfm's geek porn then?
Their 27TB array.
Unfortunately they prove that its not the size of the disk array that matters its what you do with it. It seems lastfm crew have absolutely no idea how to scale their system beyond throwing some more memory/disk space/CPU power at it
http://www.flickr.com/photos/lastfm/2266368081/ - 128gb ram
http://www.flickr.com/photos/lastfm/2267157434/in/set-72157603915907717/ - a quad quad core xeon...
Oh, and the power supply problem wasn't their fault, it was the datacentre's, basically one powersupply blew so the datacentre had to turn everything off (or run minimal systems) until it was replaced.
I actually used to design large computer installations. Probably for a bigger customer base than last.fm's too.
I'd generally have a couple of fecking big switching boxes in front to spread traffic out to more than one data centre in different locations, preferably in different countries.
This appears to be what the last.fm people do best:
Last FM have to be some of the biggest chancers out there.
They should get Shep's answerphone to run them.
[quote:58586154f9="last.fm"]So we got our new NAS box for uber storage. How long will it last? Not long probably...
Update, 4/12/2007 - Well, we only have 4.1TB left :( 85% full
Update, 16/2/2008 - It's full. :([/quote:58586154f9]didn't last too long then...
[quote:599e9af994="my-angel-rocks"]Oh, and the power supply problem wasn't their fault, it was the datacentre's, basically one powersupply blew so the datacentre had to turn everything off (or run minimal systems) until it was replaced.[/quote:599e9af994]
Eh? Every storage centre I've ever seen has two supplies to the system - which failover instantly. So how did one supply blowing having any effect on the system as a whole? If it was the supply to the supplies (so to speak) then the standard build of having two different power rings (with the redundant supply ring powered by a maintained system, which in turn is backed up by a UPS - ours is the size of a large portacabin!) would have sorted it? No need for distributed sites as you suggested steve, and much cheaper!