RAID level: the most important RAID concepts compared

Combining hard drives as a RAID has various ad­vant­ages and dis­ad­vant­ages. These cannot be gen­er­al­ised since the specific prop­er­ties of a network depend on the selected RAID level. Each level is different, depending on how many hard disks are combined and which methods are used to store the files.

We explain how RAID levels differ and what the effects of the different ap­proaches are. Our com­par­is­on of RAID levels also shows how each performs in terms of cost, re­li­ab­il­ity, and per­form­ance.

RAID levels such as RAID 5 or RAID 6 describe different ap­proaches for con­nect­ing hard disks in a network that functions as a single logical drive. When RAID tech­no­logy was first presented in 1988, the idea was to save costs by combining disks in a hardware cluster. Today, the cost factor barely plays a role and is only con­sidered when eval­u­at­ing between the various levels.

The decisive prop­er­ties in which RAID levels differ from one another today are as follows:

• Degree of re­si­li­ence
• Read speed (output rate)
• Write speed (input rate)
• Util­isa­tion of available storage ca­pa­cit­ies
• Minimum number of hard drives required

The primary goal of RAIDs is to increase the security of stored data. Failures of in­di­vidu­al hard drives are com­pensated for and data loss is avoided. To this end, in­di­vidu­al RAID levels employ different tech­niques to store files re­dund­antly. The classic procedure is, for example, the mirroring of all data whereby in­form­a­tion is du­plic­ated in a disk cluster. Al­tern­at­ively, other RAID levels rely on parity in­form­a­tion that is stored together with the user data on the in­teg­rated data carriers and – in case of a defective storage medium – helps to restore data quickly and easily.

To boost per­form­ance, many RAID levels make use of striping, a process whereby the stored data is broken down into strips and dis­trib­uted evenly across all in­teg­rated hard drives. In this way, both the write and read speed can be optimised – the degree of increased per­form­ance depends on the chosen re­dund­ancy method.

Compared to in­di­vidu­al drives, many RAID levels offer both increased re­li­ab­il­ity and improved per­form­ance. However, the following rule of thumb applies: the higher the fail-safety of a network, the weaker its per­form­ance im­prove­ment.

The maximum storage capacity users have available for their data depends on the methods used for re­dund­ancy. RAID levels that rely on mirroring data only have 50 percent of storage space available. Where parity is used in a RAID, the storage capacity increases with the number of hard disks.

The re­la­tion­ship between fail-safety, per­form­ance, storage capacity and ul­ti­mately also the cost differs from one RAID level to the next. Some ap­proaches such as RAID 0 and RAID 1 are designed for a single purpose. While RAID 0 ensures increased data through­put when reading and writing, RAID 1 focuses on the duplicate storage of files and thus only a minimal im­prove­ment in read speed is achieved (with a suitable RAID con­trol­ler).

The table below compares the prop­er­ties, strengths, and weak­nesses of common RAID levels.

For many years, RAID systems were the ultimate in saving data in a more fail-safe and per­form­ance-boosting manner. However, plenty of al­tern­at­ive tech­no­lo­gies now exist.

Multi-copy mirroring is a practical RAID al­tern­at­ive whereby – as with mirroring in RAID level 1 – several identical copies of the data are created. In contrast to a RAID system, however, these copies are on different hosts in the network and include a mode that con­tinu­ously checks the status of the data. If this recovery mechanism en­coun­ters damage or in­ac­cess­ible data, it is im­me­di­ately repaired using a copy. The number of copies is de­term­ined by the user, with each copy always taking up as much storage space as the original, which makes MCM a rather expensive pro­pos­i­tion.

Erasure Codes use al­gorithms that break down data into subsets or blocks – just like the striping method as in RAID levels 0 or 5. The in­di­vidu­al data parts can be con­veni­ently dis­trib­uted to separate storage locations. Erasure codes also have a checking mechanism that ensures the read­ab­il­ity and avail­ab­il­ity of the data. The RAID al­tern­at­ive requires ad­di­tion­al storage space of 33 percent, which makes it one of the most cost-effective options for file backup. Erasure codes are best suited to storing large data sets, as the division into blocks is par­tic­u­larly efficient here.

We’ve already touched on the fact that a RAID should never be used as an al­tern­at­ive to a backup and similarly a backup shouldn’t be mis­un­der­stood as an al­tern­at­ive to a RAID. Backups in the cloud serve the sole purpose of storing data in a separate location. Improving re­li­ab­il­ity or read or write speeds aren’t the aim of cloud storage. However, as long as hardware op­tim­isa­tion isn’t your goal, cloud backup solutions are a good option for securely storing your data.