RAID Redundant Array of Independent (or Inexpensive) Disks

RAID stands for Redundant Array of Independent (or Inexpensive) Disks. There are various types of RAID, referred to using a numbering system i.e. RAID 0, RAID 1, RAID 5 etc. The basic idea is to improve reliability and performance by using multiple disks to provide redundancy and share the workload. The most common scenarios in small server systems are RAID 0, RAID 1, RAID 5 and RAID 6.

RAID 0 consists of two identical drives. When a file is written, some data goes on one drive and some goes on the other. As both drives are being written to (or read) simultaneously, throughput is maximised. However, as bits of files are scattered across the two drives, if one drive fails then everything is lost. Also, the speed of disk drives is not necessarily the bottleneck in many servers. For these reasons RAID 0 should be avoided, as it offers no protection at all, but is described here for reasons of consistency.

RAID 1 consists of two identical drives that mirror each other. So, when a file is saved there are actually two separate but identical copies behind the scenes, one held on each drive, even though you can only see one as the mirroring process itself is invisible. If one of the drives fails, the second one automatically takes over and the system carries on without a blink. At the earliest opportunity the faulty drive should be replaced with a new one; the system is then synced it so it becomes a true copy of the remaining healthy drive in a process known as ‘rebuilding the array’. In a RAID 1 system, the total usable storage capacity is half that of the total drive capacity installed. For example, if a server has two 2TB drives installed then the total amount of usable storage capacity is 2TB rather than 4TB. 

RAID 5 uses at least three but preferably four or more drives. Data is written across all the drives simultaneously, along with what is known as parity information. The benefit of this is that the system can cope with the failure of any one single drive. RAID 5 is considered to offer a good combination of price, performance and resilience. Whereas a RAID 1 system loses 50% of the total drive capacity in order to provide resilience, RAID 5 loses only about 25%. For instance, if a server has four 2TB drives installed then the total amount of usable storage capacity is 6TB rather than 8TB.

RAID 6 uses at least four but preferably five or more drives. It is similar to RAID 5, but uses two sets of parity information written across the drives. The benefit of this approach is that the system can cope with the simultaneous failure of two drives, thereby making it more resilient than RAID 5, although it loses more capacity in order to provide that resilience. If a server has five 2TB drives installed in a RAID configuration then the total amount of usable storage capacity is 6TB rather than 10TB.

What to do? If you have a server with a single drive then the question of RAID does not arise. If you have a server with two drives you should consider using RAID 1. If you have a server with three or four drives it should be configured for RAID 5 and if you have five drives then use RAID 6. The RAID level is usually set using a BIOS option on the server.

It is important to note that RAID systems require a special controller and that these come in two types. Low-cost servers may have a so-called built-in RAID controller on the motherboard; these are referred to as ‘software RAID’ or ‘embedded RAID’ by manufacturers but are known colloquially as ‘fake RAID’ and are to be avoided as they can be prone to failures and data loss. Instead, a separate hardware RAID controller card should be purchased as this offers better performance and reliability. 

One important thing to note is that a RAID system is not a backup system. It can help prevent data loss in the event of problems but it is still important to make separate provision for backup. For instance, if the server was stolen or the premises went up in flames then the data would be lost regardless.