|RAID 0||Striping||locks all the disks into synchronization and reads and writes portions of data to different disks at the same time thereby yielding very high data transfer rates, but no improvement in access speed. There is no redundancy in the disk usage in RAID 0.|
|RAID 1||Shadowing/Mirroring||writes the same data to several disks that are synchronized but out of phase. This offers a great improvement in access speed and improved read performance , but, at best, no improvement in write data transfer rates.|
|RAID 2|| ||Error corrected one bit per disk. Offers very high transfer rates with no improvement in access time. Used occasionally in supercomputer environments -- Superseded by RAID 3.|
|RAID 3|| ||Block (usually Byte) Striping with Dedicated Parity on a separate disk.|
|RAID 4|| ||Conventional storage with parity information on a separate drive. Any performance improvement comes from the capability to do simultaneous read operations of different files.|
|RAID 5|| ||Similar to Raid 4 but with parity distributed over the data disks and always on a different drive than the data.|
|RAID 6|| ||Block Striping with two sets of Distributed Parity.|
Performance for an 8GB disk Array Disk Required for 8Gb effective storage: R0(2x4Gb); R1(2x8Gb); R2(11x1Gb), R3/4/5(10x2Gb) Read Transaction Rate relative to 8Gb non-RAID R0/2/3=Same, R1=2x, R4=4x, R5=5x Write Transaction Rate relative to 8Gb non-RAID R0/1/2/3/4=Same, R5=2x+ Transfer Rate relative to 8Gb non-RAID R0=Same, R1=2x, R2=8x, R3/R4=4x, R5=5X(read)/2x+(write)
RAID is generally implemented using SCSI interfaces but as of early 1999, IDE RAID controllers are starting to appear.
Return To Index Copyright 1994-2002 by Donald Kenney.