RAID 6: Storage tech­no­logy to minimise data loss

Combining multiple hard drives into a single logical drive isn’t exactly a new approach. Over the years, various setups have proven suc­cess­ful, and become stand­ard­ised as RAID levels (Redundant Arrays of In­de­pend­ent Disks). One approach that is rarely used, but is relevant thanks to its high fail-safety, is RAID 6. Find out what makes RAID 6 stand out, its strengths and weak­nesses and when you should use it.

RAID 6 is a data storage approach that combines four or more hard drives into a single logical drive. Compared to in­di­vidu­al data carriers, it boosts re­li­ab­il­ity of the system and read speeds. The basis for this is the com­bin­a­tion of striping and parity, that also forms the basis of RAID level 5. It comes as little surprise then that RAID 6 is often referred to as ‘RAID 5 expansion’.

RAID 6 systems implement the ‘striping’ approach in a classic manner. All data is divided into blocks and dis­trib­uted evenly to the par­ti­cip­at­ing hard disks. This gives users the option of accessing several disks at the same time and reading the sub-blocks of a data strip in parallel.

When it comes to parity, RAID 6 differs from other levels: the system always saves two sets of parity in­form­a­tion. In that way, as­so­ci­ated data can be restored if one or two disks fail. For this purpose, a RAID 6 system can op­tion­ally use the XOR logic or a mix of XOR logic and multi-bit error cor­rec­tion using Reed-Solomon code. The latter is also required to transmit tele­vi­sion signals according to the DVB standard, where it improves the bit error rate of the received signal.

The total storage capacity of a RAID 6 drops sig­ni­fic­antly compared to that of in­di­vidu­al drives. The space available for user data can be easily cal­cu­lated using the following formula:

For example, with four 1GB hard disks, only 50 percent of their potential memory would be available to store user data. However, as the number of disks increases, this re­la­tion­ship between capacity and parity improves.

Compared to RAID 5, the extended RAID 6 has one decisive advantage: parity in­form­a­tion to recover lost data is saved in duplicate. Du­plic­ated parity data is a more efficient way of creating re­dund­ancy, and also ensures higher re­li­ab­il­ity. In the RAID 6 system, up to two hard disks can fail at the same time without en­dan­ger­ing the operation of the system.

The ad­vant­ages of RAID 6 are also closely linked to the strengths of RAID 5. The pos­sib­il­ity of parallel access ensures a better through­put rate when reading data. Because of double parity, this advantage is weaker compared to level 5 systems. The ratio between the capacity for storing user data and no storage capacity (for parity) is weaker compared to RAID 5. However, when five hard disks are installed in a RAID 6, the system proves more resource-efficient and can be scaled with every ad­di­tion­al hard disk.

Compared to single drives, the sig­ni­fic­ant reduction in potential storage space is one of the system’s major dis­ad­vant­ages. This is par­tic­u­larly evident in a system of four disks where just 50 percent of storage capacity is available. The write per­form­ance is another dis­ad­vant­age in a RAID 6 network. If data is saved on the linked hard disks, parity is cal­cu­lated and dis­trib­uted twice each time. This minimises the write rate, which is evident when re­build­ing the system, i.e., when in­teg­rat­ing new hardware as a re­place­ment for defective copies.

Because RAID 6 acts like a buffer between two possible hardware failures, its potential to store large amounts of data in a fail-safe manner long-term is huge. Ideal ap­plic­a­tions are therefore server systems where large amounts of data are archived. Database or trans­ac­tion servers, which typically use RAID 5, can also benefit from a RAID 6 storage structure if the security of the data is the focus and a minimised write speed is com­pat­ible with the re­quire­ments of the re­spect­ive ap­plic­a­tion.

RAID 6 is a fail-safe al­tern­at­ive to the more widely used RAID 5. But re­dund­ancy through parity isn’t the only option. Other standards such as RAID 1 and RAID 10 store data in a mirrored manner, i.e., always in duplicate. The latter approach combines two RAID levels with one another, in that it not only includes the mirroring tech­no­logy of RAID 1, but also dis­trib­utes the data according to RAID 0 to all in­teg­rated hard drives.