140 likes | 324 Views
RAID: Redundant Array of Inexpensive Disks. Supplemental Material not in book. RAID. Redundant Array of Inexpensive Disks (or Independent Disks) proposed with five levels (RAID 1-5) in 1987 RAID employs two concepts: Striping Parity
E N D
RAID: Redundant Array of Inexpensive Disks Supplemental Material not in book
RAID • Redundant Array of Inexpensive Disks (or Independent Disks) proposed with five levels (RAID 1-5) in 1987 • RAID employs two concepts: • Striping • Parity • Concepts allowed for larger logical disks, greater speed and / or higher reliability • Some initial RAID technologies are defunct (RAID 2), and other new variations have been defined (RAID 6, 0+1, 1+0, 5+0, etc.)
Striping • By using multiple disks, and writing data to every disk simultaneously, can improve read and write performance • Having each disk on their own channel can further improve performance • Multiple disks striped together appear to the operating system as one larger logical drive • Advantages: speed and large volume size
Parity • By using multiple disks, and writing a parity bit for every bit written, the system can recover from a disk failure • Advantage: Provides a level of fault tolerance Good examples for RAID • http://www.acnc.com/raidedu/10 • http://www.lascon.co.uk/hwd-raid.php
RAID 0 • Not part of original RAID specification • Only consists of striping (no parity) so not true RAID • Can provide fastest read and write times • Has weakness in a single drive failure means the entire RAID is lost • Not recommended for important, changing data • Good for fast access to large volumes of backed up, unchanging data
RAID 1 • Original RAID specification for 2 disks in a mirror • No parity – as a full mirror, parity not required • No striping • Provide fast read, slower write times • Can recover from loss of single drive • Usable data space: 1 disk out of 2
RAID 4 • Original RAID specification for multiple disks in a stripe, with one disk maintaining parity • Provide very fast read, slower write times • Can recover from loss of single drive • Usable data space: n-1 disks out of n
RAID 5 • Original RAID specification for multiple disks in a stripe, with parity distributed across all disks • Provide very fast read, better write times than 4 • Can recover from loss of single drive • Usable data space: n-1 disks out of n
RAID 6 • Not in original RAID specification. For multiple disks in a stripe, with dual parity distributed across all disks • Provide very fast read, slower write times • Advantage over RAID 5 is dual parity can allow for loss of 2 drives. • Usable data space: n-2 disks out of n
RAID 50 • Not in original RAID specification. Known as nested RAID. A set of RAID 5 stripes, then brought together into a larger stripe. • Provide very fast read, slower write times • System can lose multiple disks, as long as only one disk in an individual RAID 5 stripe is not lost. • Usable data space: n-s disks out of n, where s = number of RAID 5’s present in the stripe
RAID 10 & Raid 0+1 • Not in original RAID specification. Known as nested RAID. A set of mirrored disks, then brought together into a stripe. • Provide very fast read, slower write times • System can lose multiple disks, as long as both disks in a mirrored pair are not lost. • Usable data space: n/2 disks out of n (n is even) • 10 is stripping mirrored sets • 0+1 is mirroring stripped sets – e.g database with remote real time backup copy
RAID 5 1 0 1 01P 1P 0P 0P 0 0 1 01 0 0 0 ?P ?P ?P ?P ? ? ? ? 1 0 1 01 1 1 1 Parity Calculation • If sum of non-parity bits is even, then 0 • If sum of non-parity bits is odd, then 1 • Knowing this, if a disk is lost its bits can be determined from the remaining bits and the parity bit
RAID • Software RAID • RAID can be implemented in software • Inexpensive • Lower performance, as CPU handles RAID management • In Windows, to create a software RAID, disks must first be dynamic • Hardware RAID • RAID can be implemented in hardware • More expensive • Better performance, as CPU offloads RAID management to the RAID controller • Hot Spares • RAIDs can define one or more hotspares • Hotspares are disks that can be automatically pulled into the array when a disk fails
COMMON RAID These RAIDs are commonly used: • RAID 0 (striping for large volumes and speed) • RAID 1 (mirror for full copy of system) • RAID 5 (stripe with parity, for larger volumes with redundancy) • RAID 6 (when more redundancy is desired) • Becoming more common as storage cost is falling • RAID 10 or 0+1 (for extensive redundancy coupled with high volume) • RAID 50 (for very large volume arrays with redundancy) Everything else is almost never used!