RAID Type Comparison and Recommendations

A redundant array of independent disks (RAID) is a technology that can be used to provide increased storage functions and reliability through redundancy. The following table provides a brief description of the common available RAID types and usage recommendation for CES.

Type	Brief description	Recommendation
RAID 0	Data striping without redundancy. Single disk failure destroys the array. High performance.	Not recommended since it is not fault tolerant
RAID 1	Mirroring. Provides better read performance than a single disk. Is fault tolerant.	Recommended for a single slice
RAID 10	Hybrid (or nested) RAID that is a stripe of mirrors (a RAID 0 of RAID 1). Performance is high for reads and writes. Is fault tolerant, as long as a mirror does not lose all its disks.	Recommended for two index slices
RAID 0+1	Offers a slightly better performance but is slightly less fault tolerant than RAID 10.	Recommended
RAID 5	Data striping with block level parity. Requires all drives but one to operate. Drive failure requires replacement. Read performance is adequate, but write performance is too low to be used in an indexing context.	Not recommended
RAID 1×n	A set of mirrors, one for each slice, when two slices are needed. Write and read performances are better because of the way CES evenly splits I/O operations between the RAID arrays. Basically, CES does the load-balancing instead of the controller. More complex configuration.	Recommended to achieve better performances with the same number of drives