Note: Many topics at this site are reduced versions of the text in "The Encyclopedia of Networking and Telecommunications." Search results will not be as extensive as a search of the book's CD-ROM.

RAID defines techniques for combining disk drives into arrays. Data is written across all drives, which improves performance and protects data. The alternative is to use one large drive, which does not have the performance benefits of an array and is a single point of failure.

A RAID appears as a single drive. Data is written evenly across the drives by using a technique called striping. Striping divides data over two or more drives, as shown by the crude example in Figure R-2. The figure shows characters for clarity, but data is usually written in blocks or sectors on each drive. A data file that might take 4 seconds to write on a single drive can be striped to four separate drives in 1 second. Likewise, disk reads are improved because there is a speed advantage in simultaneously reading data from four separate drives.

One form of RAID (level 3, as discussed in the following list) provides redundancy that protects against the failure of one disk in the array. Parity information is generated from the data written to each of the RAID drives, and that parity information is written to a backup drive. If one drive in the array fails, the parity information can be used to rebuild the information that is not available due to the failed drive. However, this parity technique does not provide protection if multiple drives fail. Therefore, some vendors have come up with their own redundancy schemes. Some examples can be found at the Advanced Computer & Network Corp. Web site listed on the related entries page.

Most RAID systems allow hot replacement of disks, which means that disks can be replaced while the system is running. When a disk is replaced, the parity information is used to rebuild the data on the disk. Rebuilding occurs while the operating system continues handling other operations, so there is some loss of performance during the rebuilding operation.

RAID levels are outlined in the following list. As mentioned, other levels of RAID have been developed, but some are proprietary. Additional information can be found at the Web sites listed on the related entries page. Note that RAID levels 1, 3, and 5 are most common, while RAID levels 2, 4, and 6 are rarely implemented in commercial products.

• RAID level 0    Data is striped over several drives, but there is no redundant drive, so this is not a true fault-tolerant configuration.


• RAID level 1    This is a mirroring solution. Data is written in blocks to two separate drives simultaneously.


• RAID level 2    This level provides data striping at the bit level over all drives in the array. Additional drives are used to store Hamming codes. Error-correction algorithm reconstructs data from the codes, so mirrored drives are not necessary.


• RAID level 3    Data is striped to multiple disks in blocks, and parity information is generated and written to a single parity disk. The information on the parity disk can be used to reconstruct data.


• RAID level 4    This level is similar to RAID level 3, except that data is striped in disk sector units rather than as bits or bytes. Parity information is also generated.


• RAID level 5    Data is written in disk sector units to all drives in the drive array. Error-correction codes are also written to all drives. This level provides quicker writes because the parity information is spread over all the drives, rather than being written to a single parity drive.


• RAID level 6    Similar to RAID level 5, but with added fault tolerance. A second set of parity information is written across all the drives. This is equivalent to double mirroring. This level may be more fault tolerant than necessary and has poor performance.


• RAID level 10    This is basically RAID level 1+0, where striping is applied across multiple RAID 1 pairs.


• RAID level 15    This is RAID level 1+5, where two complete RAID 5 systems are mirrored for added fault tolerance.

In the event of a failed disk, the most basic systems will require that the entire array go offline until the data on the replacement disk is rebuilt. As mentioned, some systems support hot replacement so that a disk can be replaced while the system continues to run. Data is dynamically reconstructed while the system remains online.

The Adaptec Array Guide provides everything you need to know to understand and build storage array systems. The Web site is listed shortly.

Copyright (c) 2001 Tom Sheldon and Big Sur Multimedia.
All rights reserved under Pan American and International copyright conventions.