zeropage is © by the L13-Crew, all part of TUI-NET

RAID: Redundant Array of Inexpensive Disks

• see also: RAID unter Linux - aber welches?
• see also: wikipedia.org
• Sometimes also called "Redundant Array of Independent Disks"
• in contrast to: SLED - Single Large Expensive Disk
• RAID offers different kinds of architectures, called RAID level
• There is a study called "A Case for Redundant Arrays of Inexpensive Disks (RAID)", which defines 5 levels
• This paper was published in 1987 at the University of California Berkeley by David A. Patterson, Garth Gibson and Randy H. Katz
• the number of a RAID level does not stand for its quality level in the sense of "the higher, the better"
• parameters influenced by RAID level: costs, capacity, security, write-speed, read-speed
• RAID level 0 is defacto only an AID, since there's no redundancy
• RAID level 10 is a combination of the levels 0 and 1
• RAID levels other than the ones standardized are common (ab)use in marketing scenarios
• nowadays, RAID levels 2, 3 and 4 are seldomly used
• JBOD is not really part of the RAID technology; it's simply the abbreviation for "Just a Bunch Of Disks" and is sometimes also referred to as RAID level 0 in spanning-mode

terms

ECC Error Correction Code member disk a disk drive that is part of an array mirror duplicate data :-) stripe data-distribution over the disks of an array stripe size size of one elementary addressable data block of an array strip size size of the part of one stripe which resides on a single disk (in contrast to the whole array) rank the group of member disks that are related to a logical drive

RAID levels

• c: capacity of a single medium
• m: number of media for data, natural number, even, m ≥ 2, max will be limited by your RAID implementation
• n: number of media for data, natural number, may be even or odd, n ≥ 1, max will be limited by your RAID implementation
• r: number of media for redundancy, r ≥ 1

RAID level # of disks capacity (netto) capacity "lost" description single disk 1 c 0 no redundancy and no speed-up JBOD Just a Bunch Of Disks n n * c 0 disks are simply concatenated; may even be of different sizes RAID level 0 block striping m m * c 0 at least 2 disks, no redundancy, simply making a virtual large disk out of all member disks. May result in higher I/O-performance if block striping instead of the (older) spanning-only method is used (see JBOD above therefore) RAID level 1 drive mirroring or duplexing n + r n * c r * c at least 2 disks, all data is written to both disks, may improve read-performance and may degrade write-performance. Can be used in combination with multiple controllers (typically called duplexing) RAID level 2 striping with ECC n + r n * c r * c data is distributed in single bits over the member disks, with ECC RAID level 3 byte striping with parity n + 1 (n ≥ 2) n * c c parity information is stored on a dedicated disk RAID level 4 block striping with parity drive n + 1 (n ≥ 2) n * c c like RAID level 3, but not on block level instead of byte level RAID level 5 block striping with distributed parity n + 1 (n ≥ 2) n * c c like RAID level 4, but parity is distributed over all disks RAID level 6 block striping with 2 distributed parities n + 2 (n ≥ 2) n * c 2 * c like RAID level 5, but with redundant parity information RAID level 7.p block striping with p distributed parities n + p (n ≥ 2) n * c p * c like RAID level 6, but with (n-1)'th redundancy of parity information RAID level 0+1 mirrored stripeset m * 2 m * c m * c RAID level 10 striped mirrors 2 * (n + r) 2 * n * c 2 * r * c RAID level 50 striped distributed parity 2 * (n + 1) (n ≥ 2) 2 * n * c 2 * c RAID level 51 mirrored distributed parity 2 * (n + 1) (n ≥ 2) n * c (n + 1) * c RAID level 60 striped distributed double parity 2 * (n + 2) (n ≥ 2) 2 * n * c 4 * c NRAID tbd tbd tbd tbd