A fault-tolerant (redundant) array can continue operating after experiencing a failure
of one (or sometimes more) of its drives. When this happens the array enters a degraded
state until the faulty drive is replaced and its contents are rebuilt.
Running in a degraded state has an important impact on both the performance and subsequent
fault tolerance of the array; see this full
discussion for more.
Some RAID levels handle degradation more "elegantly" than others; some can in
fact tolerate the loss of more than one drive, while others cannot handle the loss of even
one. While in a degraded mode, some RAID levels suffer much more performance loss than
others; generally, the more complex the RAID level, the more its performance decreases
when the array is degraded (simple mirroring shows little loss of performance, while
striping with parity shows much more). Also, alternate ways of
implementing multiple RAID levels can be surprisingly different in terms of the number
of lost drives they can tolerate, and also in the impact on performance when rebuilding
after a fault.
Next: Performance
