RAID ARRAYS

What is RAID?

RAID is a system devised for safeguarding data against hard-disk failure. It is an algorithm for "Redundant Array of Inexpensive Disks". The purpose is to not only safeguard data but to increase the MTBF (Mean Time Between Failures) for the host system.

A system with a RAID array will continue to function when a disk fails, allowing time for a replacement disk to be sourced and fitted. In mission-critical environments (busy webservers, large corporate fileservers etc) a Hot-Swap system is often used which allows the disks to be replaced without even shutting down the host computer. Whilst this technology is getting cheaper, it is still largely an unnecessary luxury for the SME.

RAID arrays are typical in mid and high-end small business servers as well as the 'big-iron' used in larger companies. There are many variants of the RAID array but level 1 and 5 are the most common.

RAID level 1 (RAID-1) is mirroring of disks – i.e. identical data is written to two disks so if one fails the other still retains the data and continues to function. This is inefficient storage however as you have to buy double the capacity that you require. RAID-1 can be achieved at operating system level, without the need for a RAID controller card. Although this is not generally recommended, it is a cheap option and providing the host computer has ample power, works adequately. A dedicated RAID card will relieve the processing from the central processors of the computer and is thus far more effective and offers better performance.

RAID levels 3-5 works slightly differently in that data is distributed evenly over multiple disks and a parity check (checksum) of the data is written to an additional disk – RAID-3 (byte-level) & RAID-4 (block-level) or distributed between the disks (RAID5). If one of the disks fails, the missing data can be calculated and reconstructed on-the-fly by the controller card using the parity information. Thus the system will continue to function, albeit with a performance cost. When a faulty disk is replaced, the controller card will rebuild the missing data onto the new drive whilst still performing its core role of reading and writing the array. This type of RAID obviously requires at least 3 disks, but the more are in the array, the more efficient the system. I.E. if you have an array of 6 10GB disks, one sixth will be parity information whereas if you have three, a third will be parity.

Systems with no fault-tolerance are often referred to as RAID 0 arrays (i.e. none!)

Other configurations are possible, for example two RAID-5 arrays could be mirrored or each component drive could be individually within a RAID-5 array but these types of configurations are less common and rarely necessary for the SME.