What is RAID (Redundant Array of In­de­pend­ent Disks)?

RAID improves the fail safety of data storage devices. Developed for use in HDD hard drives, it’s still used today in server en­vir­on­ments. What exactly does the structure of a RAID system look like and what are the dif­fer­ences between the in­di­vidu­al levels?

Over the coming years, RAID was stand­ard­ised and developed – its suit­ab­il­ity to server ap­plic­a­tions in­creas­ingly coming to the fore. As a result, it became less about saving money and more about ex­chan­ging hard disks without causing op­er­a­tion­al errors. It’s all in the name: Redundant Array of In­de­pend­ent Disks. The RAID tech­no­logy is spe­cific­ally tailored to the prop­er­ties of classic HDD hard drives. Modern SSDs can tech­nic­ally be bundled, but at the loss of per­form­ance and lifespan due to the missing TRIM func­tion­al­ity in RAID.

RAID systems are still important com­pon­ents in server en­vir­on­ments today. Their most important role is the re­dund­ancy of the stored data. This shouldn’t be confused for the concept of data backups. As part of a server structure, RAIDs ensure that the failure of a single hard drive doesn’t have severe con­sequences, because its data is stored elsewhere in the RAID system. Other ad­vant­ages of RAID are an increase in storage capacity and faster read and write speeds when accessing the hard disk space.

How in­di­vidu­al storage media of a RAID system interact, and which function a network should ul­ti­mately fulfil in a server network varies widely. However, there are various stand­ard­ised setups that are defined in so-called RAID levels. In addition, a dis­tinc­tion is made between software and hardware RAIDs, depending on whether the in­ter­ac­tion of the network is organised on the software or hardware side.

Cat­egor­isa­tion into hardware and software RAIDs can lead to the wrong im­pres­sion of what these two types of hard disk drives are all about. Both variants require software to operate – the terms only refer to the type of im­ple­ment­a­tion.

With hardware RAID, the or­gan­isa­tion of the in­di­vidu­al storage media is performed by special, high-per­form­ance hardware, also known as a RAID con­trol­ler. This con­trol­ler is installed either in a computer or a disk array that also houses the hard disks. Disk arrays are commonly used in data centres, whereby the external systems are often referred to as DAS (Direct Attached Storage), SAN or NAS. The great advantage of hardware-based or­gan­isa­tion of RAIDs is excellent per­form­ance in the form of high data transfer rate.

In a software RAID, the storage quota is managed by software that runs directly on a host’s CPU, also referred to as host-based RAID system. Common operating systems such as Windows (as of NT) or Linux dis­tri­bu­tions include the necessary com­pon­ents. Compared to the hardware al­tern­at­ive, a software RAID is set up much faster and cheaper. Dis­ad­vant­ages are high CPU for the host and the lack of platform in­de­pend­ence. Since disk access cannot be regulated as elegantly as with a RAID con­trol­ler, per­form­ance also tends to be worse.

The manner in which hard disks are combined in a RAID is called a ‘level’. However, this can cause some mis­un­der­stand­ings, because hard drive setups don’t build upon one another in levels. The in­di­vidu­al levels are not connected and only char­ac­ter­ise the different struc­tur­al ap­proaches and functions of the RAID. Common levels include RAID 0, RAID 1, RAID 5 and RAID 6. Com­bin­a­tions of two RAID levels are also possible. RAID 10, for example, des­ig­nates a RAID 0 system that has been combined from several RAID 1 systems.

Strictly speaking, hard disk groups under a RAID 0 label don’t count as RAID systems at all, as they don’t rely on re­dund­ancy for storage. The only purpose of RAID 0 is to ac­cel­er­ate access to data by combining two or more hard drives into a single logical drive. To this end, data is evenly dis­trib­uted across the in­di­vidu­al data carriers in suc­cess­ive blocks that are called stripes. That’s why RAID 0 is also known as “striping”. While the network provides more storage capacity and a higher through­put rate, it lowers security: if a hard drive fails, all data is lost. You can find out more about striping in our detailed guide to RAID 0.

RAID level 1 is also known as ‘mirroring’. Here, all hard disks always have the same data status providing for excellent re­dund­ancy and data safety in case of failing storage media. As such, the capacity of the RAID is always as high as the capacity of the smallest hard drive installed. The write speed in a RAID 1 is as fast as that of a single drive. By con­nect­ing the com­pon­ents to their own channels, such as SATA, reading speed can be doubled. Find out more about ‘mirroring’ as a storage method in our article on RAID 1.

RAID 5 describes a network of three or more hard disks, the number of which is typically uneven – three, five, seven, etc. The storage concept uses the striping of RAID 0 and dis­trib­utes the data in blocks across the various data carriers. Together with the data blocks, parity in­form­a­tion is evenly dis­trib­uted across the in­teg­rated hard drives, which can be used to restore lost data if a storage medium fails. Thus, RAID 5 ensures a higher read speed and also more security than a single drive. Due to the constant need to re­cal­cu­late the parity blocks, the writing speed is com­par­at­ively slow. Read more about the concept in our separate RAID 5 article.

RAID level 6 follows a similar approach to RAID 5: data is dis­trib­uted evenly and in blocks to the in­teg­rated storage com­pon­ents, and parity in­form­a­tion makes for higher security. However, recovery data is generated in duplicate, which means the RAID type can cope with the sim­ul­tan­eous failure of up to two hard disks (when a minimum of four are installed). The network therefore offers a high level of data security and good read access. Since cal­cu­lat­ing the parity blocks is even more time-consuming than with RAID 5, the write speed is slower. In our guide on RAID 6 we discuss the strengths and weak­nesses of striping with doubly dis­trib­uted parity in­form­a­tion.

RAID 10 or RAID 1+0 combines the features of RAID level 0 and RAID level 1: an increased data through­put rate and higher data security. For this purpose, several RAID 1 systems are combined in a RAID 0 array – at least four hard drives are required. Find out when this com­bin­a­tion is a good idea and what its dis­ad­vant­ages are in our article on RAID 10.

There are a few things to consider when setting up and operating a RAID system. Firstly, you need to decide what type of network you want to set up. For example, if you’re only looking to increase data through­put, you could use a level 0 system or al­tern­at­ively opt for an SSD. If you want to boost data security, you could choose between mirroring (e.g., level 1) or storage with parity in­form­a­tion (e.g., level 5).

When selecting hard drives, choosing identical models is prefer­able. In many RAID setups, the maximum storage volume depends on the smallest disk, which is why storage potential may be lost when mixing hard drives of different sizes. More im­port­antly, rely on hardware such as NAS that are designed for endurance and a longer life. The size of the data carrier also plays an important role when defective hardware is replaced later or when the RAID is enlarged: new or added com­pon­ents must be at least the size of the smallest existing or defective data carrier.

Here’s an important reminder when using RAID systems: while the in­ter­ac­tion of hard drives improves the security of the stored data through re­dund­ancy, it cannot replace a good backup solution.