I am building a personal fileserver. I am building it around an ABIT AT7 motherboard, which has the standard 2 IDE controllers and the HPT374 controller (4 IDE raid channels) onboard as well.
I have the money and motivation to FILL it with 250gb drives.
What I have planned is:
IDE0 master: 250gb
IDE0 slage: 250gb
IDE1 master: 250gb
IDE1 slave: 250gb
RAID0 master: 250gb
... you get the idea.
I'm going to set IDE0 master aside for the system &c, and then software RAID 5 the other 11 disks together into one massive array: 10 disks and a hot spare.
Does anyone see a problem with this?
What class processor will I need to calculate the overhead on an 10-disk array? will I run into throughput problems, or will the network interface be the bottleneck?
Bad idea I say. Firstly the idea of a RAID5 array is security, making software RAID5 is probably the least secure implementation of this, especially over several controllers. Second, do you really need 2TB or storage space? If you do, far better to get a RAID5 controller such as the Highpoint RocketRAID 464 (PATA) 1820A(SATA) or Promise FastTrack S150 SX4-M. These will set you back around $200. Then get 400GB drives and hook 5 of them up, a much better implementation, definitely faster and more secure if you ask me. Using less drives but bigger makes the array more secure becuase 10 drives gives you 10 times the risk drive failure. The size of the array doen't increase the amount of parity calculations necessary to write data. That's the same no matter how many disks you have.