RAID 1 vs RAID 5

Recently, I had to decide between RAID 1 and RAID 5 for a server deployment related to a mission critical application. This app had a considerable amount of writes compared to reads.

While reading up on RAID I came across this good site (http://en.wikipedia.org/wiki/RAID)

Although I can repeat much of the information listed there and on many other sites, I think a summary would do much better:

RAID 1 is a mirror configuration. It requires a even number of HDDs and data is written across all the drives. If a drive in the mirror fails the other available copy is used. RAID 1 provides 1xHDD size of storage in the RAID configuration. If 2 1 TB drives are connected in RAID 1 configuration, it will result in 1 TB drive. Because of replicated data, RAID 1 provides 2x read performance and almost 1x write performance.

RAID 5 is a stripping configuration with parity. It requires odd no. of drives with a minimum of 3. Data is stripped in chunks across drives. One drive is used for parity and if a drive fails the parity data is used to regenerate the missing data. RAID 5 provides (n-1) HDD size space in the configuration. With 3 drives of 1 TB in the configuration, RAID 5 offers 2 TB storage capacity. RAID 5 offers (n-1)x read and write performance.

Biggest disadvantage with RAID 5 is that performance reduces when one drive fails until it is replaced and the data rebuilt. On the other hand RAID 5 offers better storage capacity usage in the array.

Another disadvantage with RAID 6 is that software RAID will have a huge impact on CPU usage. Hence, hardware RAID 5 is recommended.

For most purposes RAID 1 should be used unless storage capacity is important at the expense of performance degradation during failure.

I chose RAID 1 for this server and it has been performing well.