Originally posted by: Phil
Here's a link to a 5-bay hot-swap SATA caddy, we use them at work and they're very reliable. One thing I should mention is that it has a maximum transfer rate of 150Mb/sec - I don't know if this is per-drive or per-caddy (i.e. either 150Mb/sec per drive, or 150Mb/sec for all 5 drives). If it's the latter, then this will impede performance, but you'll be hard-pressed to find a regular desktop motherboard that has PCI-X slots, so I wouldn't worry about it too much
As for auto-rebuild; yes, if the controller supports it then it should start rebuilding automatically. It's
highly recommended to keep the GAM software for the card loaded (array monitoring software), as not only will this inform you of any problems, it can perform maintenance on the array, such as displaying bad block lists, adding/removing hot spare drives, etc.
You'll definitely be wanting a UPS. APC make great UPSs, but they're expensive. Belkin-made units are also pretty good.
Low-noise cooling is going to be an issue. Any half-decent 5-bay caddy will have a 60mm/80mm fan mounted in the front (see the link I posted above for an example), and these tend not to be the quietest of fans in the world - they're designed to keep drives cool, and that's it.
A good bet might be to look at an Aerogate fanbus controller- as long as it's set up correctly, it will spin the fans up and down depending on the temperature readings it receives. For example, mine spins the rear exhaust fan up if the graphics card probe reads above 65C, and spins it back down to 1400rpm (virtually silent) when it cools back down. Recommended
[Edit] The LSI controllers are good performers and reliable - definitely a good choice.
[Edit2]
This UPS should do you fine, 1500VA which gives ~800W for 4 minutes - long enough for a shutdown. Half-load should give you 12 minutes runtime, and that's roughly what you're aiming for.
[Edit3] One more thing - I cannot recommend strongly enough for you to keep a spare drive around should one of the drives in the array fail. If a drive fails, then you're skating on thin ice until it's replaced
and the array has completed a rebuild. The array will keep running if a drive dies, but you should swap the failed disk, stat, unless you enjoy running the risk of losing your data
This is why a large proportion of the RAID systems we sell at work come in a 3+1, 4+1 etc setup - 3 or 4 drive in the actual array, and one more as a hot-spare; i.e. one that will kick in automatically and begin a rebuild without any intervention. Once the dead drive is replaced, then it becomes the new hot spare.