Sunday, July 4, 2010


Most Common Flavors Of RAID | RAID Levels

With the advent of massive databases and online storage needs measured in terabytes, RAID has become the standard way to create and manage the huge disk farms required to store all that data.

RAID comes in many flavors; we've covered the most common:

This version of RAID storage concatenates multiple physical disks together to create a single virtual drive. Data may be striped across the disks to improve performance. This form of RAID is great for managing large virtual storage areas with some performance improvement, but offers no additional reliability.

RAID 1 is another name for mirroring, or creating a realtime copy of every physical device. Data written to the mirrored pair is copied to both physical devices. If one of the devices fails, the other picks up the slack so that the host system is never aware that a drive has gone bad. In its pure form, RAID 1 provides the ultimate in redundancy, but offers no performance improvements or virtual device management.

RAID 0+1
Combining RAID 0 and RAID 1 yields the best of both worlds: complete mirroring for redundancy with striping and concatenation for large volume management and performance improvement. The only down side is cost, since RAID 0+1 (like RAID 1) needs two bytes of raw storage for every byte of data stored.

RAID 3 reduces the cost of redundant storage by using parity instead of mirroring to protect against disk failure. Drives are grouped into sets; one drive is designated the parity drive for the set and contains parity data computed from the other drives. If any single drive fails, the data stored on it can be recreated from the parity data. Costs are reduced, since fewer drives are needed to implement redundant storage, but performance is hindered by the parity computation and the bottlenecks caused by a single dedicated parity drive.

RAID 5 is a refinement of RAID 3 that uses the concept of parity but distributes the parity data across all the drives in the RAID set. Bottlenecks induced by the parity drive are eliminated and costs are still reduced, but performance can still be a problem.

In exploring these different RAID models, we gave little consideration to the physical implementation of the storage subsystem. Now that we understand how these different RAID types work in theory, we need to drill down one layer into our storage subsystem and see how to implement RAID in the real world.


About bench3 -

Haja Peer Mohamed H, Software Engineer by profession, Author, Founder and CEO of "bench3" you can connect with me on Twitter , Facebook and also onGoogle+

Subscribe to this Blog via Email :