Originally Posted by Legionfalls
I want 10 TB in total, meaning 5 drives @ 2 TB each. And if I understand it correctly, Raid 5 is the most suited to my needs. Speeds of Raid 0, combined with the reliability of always having a hot spare drive on the array- automatically kicking in if one does happen to fail.
I am considering running some Linux distro, it's what I wanted to do, gain some experience, knowledge, etc. If required/easier for my needs, I can always go the Windows way, what with WHS 2011 coming soon (probably not before I get my hardware, but one can hope.)
RAID5 is not RAID0 with a hotspare. And it doesn't offer the speed benefits in the same way either. I think you have a little more reading to do before you drop all this cash on your new server...
RAID0 is a simple striping of your data across multiple drives. This gives speed benefits as all the drives can process data (somewhat) concurrently, so transfers can go faster.
RAID5 is a little tricky to explain quickly or simply, but here goes...
With RAID5 you again stripe your data across your drives. However this time you spread your information across all but 1 of the disks (unlike RAID0, which uses them all). Now you then put each stripe of data into a fancy algorithm that calculates what is known as parity for you. This parity data is written onto the last of your drives (and it's location is then rotated between your drives, so each drive contains both data and parity in a round-robin type arrangement). The way this parity is calculated is very clever - if you know the parity plus all but 1 of the original bits of data, you can reconstruct that original data using some math. So if a single drive dies you can recreate the data that was on it, meaning you don't lose data in the case of a drive failure. If a drive fails in RAID0, everything is gone. However this parity takes up space on your array, so you lose the equivalent of 1 drive's worth of capacity. With 5 2TB drives you would get 8TB of useable space, so for your 10TB capacity you will need 6 drives.
Now a hotspare is something different - but you may wish to use one. A hotspare is a drive that is sitting unused on the controller, waiting for another drive to fail. When something happens to one of the drives in the RAID5 array, the controller will automatically start to use this spare drive to recreate the failed drive. This is to minimise the amount of time the array is left in this degraded (damaged) state - if a second drive dies then all the data will be lost. This obviously costs an extra drive - so 7 drives for 10TB.
Another option is RAID6. This is like RAID5, but now you do 2 fancy bits of math on the data to give 2 different, redundant parity bits. Now you can reconstruct your data even if 2 drives fail together. It again costs another drive (so 7 drives for 10TB, or 8 drives if you want a hotspare).
Regarding speed - you get somewhat of a boost from having multiple disks working together (as you did with RAID0), but now there are a few caveats. The parity data is redundant, so you lose your max sequential speed drops to roughly what you would have had in RAID0 with 1 fewer drive. But not everything works out so well - because your parity data relies on data on EVERY drive, if you write just a small change to 1 disk you then have to read the rest of the stripe, recalculate the parity and rewrite the data back again - and that is where the downside comes from - RAID5 is terrible at small writes, making it unsuitable for OS use for example.
I hope you understand the above - if not I can try again (or go to WikiPedia - the article there is reasonable, although a little incorrect in some places).
Regarding the OS choice - Linux means you can do all the RAID stuff in software. Nice and cheaply too, and you get a rock solid platform with more recovery and setup options than you can find anywhere else. But the downside is the complexity - if things go wrong it can be much harder to get your stuff back than if you went with a good hardware card. And when your array has fallen apart and you need to recover your wedding photos before your wife leaves you probably isn't a good time to start learning the intricacies of the Linux commandline...
I would suggest you have a proper think about exactly what you want your server to do, your budget, and what kind of backup/recovery strategy you want/need before you spend your money on anything.