RAID 5 is likely to be the most commonly used flavor of RAID in the industry today. It offers the best balance in performance and reliability suited to most users. Below is a diagram of the layout of blocks in a RAID 5 setup.
| | | |
0 A0 B0 C0 0P
1 A1 B1 1P D1
2 A2 2P C2 D2
3 3P B3 C3 D3
Letters represent different hard disks.
Numbers represent different stripes.
xP represents the parity data for stripe x.
One can easily see that if one disk is removed (say D), that all off the blocks on this disk can be reconstructed (1P along with A1 and B1 can generate D1, and so on for D2, D3). So, no single disk failure can cause a loss of data. Also, enterprise RAID devices tend to have hot swappable hard drives, and controllers that will properly build a new hard drive when inserted after failure. Also, you maintain excellent speed over a dedicated parity disk (as seen in RAID 3 and RAID 4, because all writes make one write to all N disks (N-1 data writes and 1 parity write). Therefor, there is no bottle neck as there would be at a dedicated parity disk.
Finally, to be useful you must use at least three drives. You still end up with effectively one whole drive dedicated to parity, though that disk is split between all disks. So if you use N disks, you get N-1 disks worth of data.
The only real disadvantages of RAID 5 are that the RAID controller must be much more complicated (read expensive), and in case of failure, the rebuilding operating is much more complicated.