• Cost of disk storage
• Data protection level required
• Performance level required
• Future expansion

When you look at cost for example you need to look at the trade off between performance and protection. The higher the protection level required the more disks are needed but can be costly in terms of disk space lost, up to 50% in the case of a mirrored system. If it is performance you require then less disks are needed but a failure could end in the loss of all data.

RAID 0 (Striping)

Requires a minimum of two disks to implement a striped disk array where data is broken down into blocks and each block is written to a separate disk drive. This level provides the best performance as you are spreading the I/O load across multiple channels and drives however there is no fault tolerance and a failure of one disk will result in all data being lost.

RAID 1 (Mirroring)

Again requiring a minimum of two disks, RAID 1 offers a cost effective way of adding a high level of fault tolerance to a system. All data is written to separate disk drives so two complete sets of data is stored within the system offering the highest level of data protection for any environment.  There is a cost disadvantage due to losing 50% of the purchased disk capacity.

RAID 5 (Striping with parity)

This level requires a minimum of three disks and one is lost to the parity element of the configuration.  Like RAID 1 data is striped across the disk drives to provide performance however for each stripe through a disk one block is reserved to hold parity data which is calculated from the other stripes on the disk. In the event of a disk failure the RAID set can rebuild on a new disk using the data held within the parity blocks on the array.  Read performance is very good in RAID 5 and there is the added benefit of data security however there is added cost to as you require one extra disk over the required usable capacity, additionally using high capacity drives can result in a long rebuild times in the event of a disk failure.

RAID 6 (Striping with dual parity)

RAID 6 uses the same principles as RAID 5 but offers dual parity to store and recover data and requires a minimum of 4 disks.  By having dual parity it can tolerate the failure of two disks in the array providing a higher level of protection than RAID 5. By having a higher level of protection RAID 6 allows the use of less reliable SATA disks for business critical data. Disadvantages of RAID 6 include the additional cost of the extra disk and like RAID 5 a long rebuild time in the event of drive failure.

RAID 10 (Striped RAID 1 sets)

This is the combination of levels 1 & 0 to offer the improved performance of striping and redundancy of mirroring. This is done by forming a RAID 0 striped array from two or more RAID 1 mirrored arrays.

RAID 50 (Striped RAID 5 sets)

This is the combination of levels 5 & 0, the striping in RAID 0 helps increase the capacity and performance without having to add disks to each RAID 5 array. RAID 50 is made up of a RAID 0 stripe across multiple RAID 5 arrays to gain the data protection benefits of RAID 5 whilst adding RAID 0 allows the arrays to be seen as one logical drive volume. By using RAID 50 a system can sustain multiple disk failures, up to one in each RAID 5, without data loss. Additionally RAID 50 dramatically improves rebuild times following a disk failure.

RAID 60 (Striped RAID 6 sets)

This is the combination of levels 6 & 0. The dual parity of RAID 6 allows the failure of two disks in each RAID 6 array and as with RAID 50 the stripping helps increase the capacity and performance. In terms of benefits again these are the same as a RAID 50 configuration.