I know the basic concepts of RAID(s) but I’m damned if I can find a clear evaluation. I’m looking for server solutions in a 15-20 user environment with a sluggish database.
The original premise of RAID was better performance using cheap disks. Gradually that evolved, in higher incarnations of RAID, to include fault tolerance (parity), which downgrades performance especially on writes, and especially with IDE (cheap) drives.
In other words, most of the performance is gone and instead you’ve got fault tolerance/redundancy.
Mirroring seems different from parity. Parity would seem to suggest that limited scope errors can be corrected by storing a limited amount of additional data on the same drives (RAID5); however, full redundancy (mirroring) requires complete duplication of the data on a separate drive.
OK, OK, the questions:
1. Does RAID[>0] really produce net performance gains?
2. Why bother with RAID if you want redundancy? Why not just mirror? At the most it saves 1 drive in 4, but adds complexity.
3. If there is a benefit to RAID (as opposed to mirroring) what is the cost?
Am I way off base here?