How to improve disk reliability ?

Disk used to be the least reliable component of the system. They still have relatively high failures and it causes loss of data and significant downtime. It takes hours to recover from a disk crash. Improving the reliability of disk systems is very important and several improvements in disk-use techniques have been proposed.
To improve speed, disk stripping uses a group of disks as one storage unit. Each data block is broken into several sub blocks, with one sub block stored on each disk. The time required to transfer a block into memory improves because all the disks transfer their sub blocks in parallel. If the disks have their rotation synchronized, the performance improves because all the disks become ready to transfer their sub blocks at the same time rather than waiting for the slowest rotational latency. This organization is usually called a redundant array of independent disks (RAID).

The simplest RAID organization, called mirroring or shadowing, just keeps a duplicate copy of each disk. This solution is costly but it is about twice as fast when reading, because half of the read requests can be sent to each disk.
Block Interleaved Parity, another RAID organization uses much less redundancy. A small fraction of disk space is used to hold parity blocks. Each bit position in the parity block would contain the parity for the corresponding bit positions in each of the 8 data blocks. If one disk block becomes bad, all its data bits are essentially erased but can be recomputed from other data blocks plus parity block.

A parity RAID system has the combined speed of multiple disks and controllers. But the performance is an issue during writing, because updating any single data sub block forces the corresponding parity sub block to be recomputed and rewritten.