How RAID 6 protection works
This topic describes how RAID 6 protection works.
The IOA determines how parity sets are formed. RAID 6 protection gives you optimal performance, capacity, and balance, so selecting any of these parity set optimizations is meaningless and will not affect the outcome of the parity set. If you choose to optimize by availability, a greater level of protection is achieved, because it allows a parity set to remain functional in the event of a single bus failure on the IOA. The parity set is formed from at least four disk units of equal capacity, with no more than two disk units attached to an individual bus on the IOA.
It is possible to include additional disk units of the same capacity in a parity set after device parity protection is initially started. You can include up to two disk units at the same time. However, if three or more disk units are present and eligible for device parity protection, the system requires that you start a new parity set, rather than include them in an existing parity set. In IBM® Systems Director Navigator for i5/OS and System i® Navigator you can view the properties of each disk unit. If the protection status of a disk unit is unprotected, it is not protected by device parity protection or mirroring and may be eligible to be included in a parity set or to be started in a new parity set. This will also be indicated by the model number which should be 050. You can also exclude disks that do not store parity data from a parity set without stopping device parity protection. You can exclude a protected disk unit with a model number of 090 because it is a disk unit that does not store parity data.
When a parity set grows, you may want to consider redistributing the parity data. For example you may begin with seven or fewer disk units, but expand to ten or more by including more disk units. When this happens, you can improve the performance on the parity set by stopping parity protection and starting it again.
A write cache is included in the IOA for each parity set to improve performance of interactive write workloads.