Recently, RAID is widely used for flash-based solid-state
disks (SSDs). Using the conventional RAID architecture with
SSDs, however, may significantly decrease data reliability
because several disk failures can be highly correlated in
SSD-based RAID systems. Because of its mechanical nature,
failures of hard disk drives (HDDs) usually occur in a random
manner; that is, a probability of more than two HDDs being
malfunctioning at similar times is very low. On the other hand,
the reliability of an SSD is largely decided by the amount of
data written to it. Since a RAID system tends to write all
incoming data to different disks equally, SSDs in the same
RAID group are likely to be worn out at similar times, which
greatly increases a possibility of multiple simultaneous disk
failures. Once several SSDs die simultaneously, it is difficult
to recover original data, regardless of the existence of parity
chunks.