RAID Comparison: Comparing RAID 0 — RAID 6
RAID storage can be configured in multiple ways. These configurations, defining how data is dispersed among multiple disks are called RAID levels. Different RAID levels exhibit different performance and reliability. The most common schemes are reviewed below.
RAID 0
Striping without mirroring. Data blocks are spread among multiple disks. Since data is accessed in parallel, it provides improved performance. As no redundant data is added, RAID 0 does not improve reliability and failure of any of the drives in the RAID-group causes a complete failure.
RAID 1
Mirroring without striping. The data is duplicated on two drives, creating a complete mirror. RAID 1 systems show improved read- with a minimal degradation of write- performance. As long as one drive in the mirrored RAID 1 group is alive, data is available.
RAID 10
Mirroring with striping. Data is written in stripes on primary disks, which are being mirrored to secondary drives. A typical RAID 10 group consists of four drives: two for striping and two for mirroring.
RAID 2
Bit-level striping with dedicated parity. Data is striped placing each bit on a different drive, with one additional disk dedicated to storing parity information.
RAID 3
Byte-level striping with dedicated parity. Data is striped placing each byte on a different drive, with one additional disk dedicated to storing parity information.
RAID 4
Block-level striping with dedicated parity. Data is striped placing each block on a different drive, with one additional disk dedicated to storing parity information. RAID 4 exhibits improved performance since each disk serves I/O requests independently allowing accessing data in parallel. The single parity drive of a RAID 4 system often becomes the performance bottleneck.
RAID 5
Block-level striping with distributed parity. Both the data and the parity information are striped across all drives in the RAID-group. RAID 5 exhibits improved performance since each disk serves I/O requests independently allowing accessing data in parallel. A single drive failure, however, causes performance degradation of the entire RAID 5 system.
RAID 6
Block-level striping with double distributed parity. Both the data and the parity information are striped across all drives in the RAID 6 group, and redundancy information is duplicated. RAID 6 exhibits improved performance since each disk serves I/O requests independently allowing accessing data in parallel. RAID 6 systems can withstand complete failure of two drives in the group. Redundant parity information provides additional time to rebuild redundancy without the risk of the information loss. Most of RAID 6 based systems still exhibit degraded performance until failed drive has been rebuilt.
|
RAID 0 |
RAID 10 |
RAID 6 |
RAID 6, initialized |
|
|
Small block read-performance |
Good |
High |
Average |
Average |
|
Large block read-performance |
Good |
High |
Good |
Good |
|
Small block write-performance |
Good |
Average |
Average |
Average |
|
Large block write-performance |
Good |
Poor |
Good |
Good |
|
Availability |
Low |
High |
High |
|
|
Drives capacity utilization |
Excellent |
Poor |
Good |
Good |
|
First-use initialization |
Fast |
Slow |
Non-standard RAID levels
RAID 7.3 and RAIDIX N+M was developed by RAIDIX in-house research lab to improve total availability and data protection.
RAID 7.3 is similar to RAID 6 but it has a higher degree of reliability for three checksums are calculated using different algorithms.
RAID N+M allows user to choose the number of disks for checksums allocation. RAID sustains complete failure up to 32 drives in the same group (depending on the number of parity disks).
Summary:
RAID 0 is typically used on systems working with data-flows. On transactional systems, where reliability is a must, RAID 6 becomes the obvious choice because data remains available even in the event of multiple drive failures.
RAID 0 and RAID 10 configurations are usually selected for RAID-groups with a small number of drives. Starting with eight drives, RAID 6 becomes the best option. RAID 6 configuration also gives a less chance for silent data corruption type of disk error.
As the amount of information that organizations need to process and disk drive sizes constantly grow, RAID rebuild times are becoming increasingly important. Ability to withstand multiple disk-drive failures makes RAID 6 the best option for systems requiring high-availability.