This level of RAID differs from mirroring and is similar to RAID-3 in that it computes and writes parity information that can be used to rebuild data in the event of a loss of one drive in the array. The difference between this and RAID-3 is that the parity information may be written to any drive in the array, removing the single-drive hot spot that is encountered with RAID-3.

With RAID-5, the data is striped across the drives and the parity information for the strips is written to an additional drive. There is, however, one large disadvantage to this, referred to as the write-penalty: Whenever a write occurs, the parity data needs to be read, re-computed, and written with the related data while the data and parity drives are locked. RAID-5 is excellent for read-only systems, such as data warehouses, and are useable for OLTP systems that do not incur a high transaction volume. RAID-5 uses only approximately 20 percent overhead (this percentage varies depending on the configuration of the drives) rather than the 100 percent overhead that mirroring incurs. Hardware-implemented RAID-5 performs much better than software implementations, and write-cache can decrease that pesky write-penalty.

As you'll see in our next tips, implementations of RAID-7 and RAID-S can improve upon RAID-5.