notes

Personal notes
git clone git://git.laack.co/notes.git
Log | Files | Refs

RAID.md (1757B)


      1 # RAID
      2 
      3 **Source:** Wikipedia :\
      4 
      5 **Definition:** Redundant Array of Independent Disks (RAID) configurations are storage device configurations that implement at least one of the following: striping, mirroring, parity.
      6 
      7 NOTE: There are additional RAID levels, but they are largely unused.
      8 
      9 ## RAID 0
     10 
     11 - [x] Striping
     12 - [ ] Mirroring
     13 - [ ] Parity
     14 
     15 RAID 0 is quite dangerous. While it benefits from striping and has no overhead cost for mirroring or parity, the loss of one drive is capable of destroying all data given that data is striped across all disks to improve IO speed.
     16 
     17 ## RAID 1
     18 
     19 - [ ] Striping
     20 - [x] Mirroring
     21 - [ ] Parity
     22 
     23 RAID 1 pairs drives for mirroring. This results in decent read performance and reliability, but poor write performance and storage capacity.
     24 
     25 ## RAID 5
     26 
     27 - [x] Striping
     28 - [ ] Mirroring
     29 - [x] Parity
     30 
     31 RAID 5 uses striping with distributed parity. This requires all but one drives to be able to operate. Additionally, RAID 5 requires at least three disks. This is a popular approach with three disks.
     32 
     33 ## RAID 6
     34 
     35 - [x] Striping
     36 - [ ] Mirroring
     37 - [x] Parity
     38 
     39 RAID 6 is basically RAID 5 except it has an additional parity block. This requires four disks, but can remain functional even if two disks fail.
     40 
     41 ## RAID 10
     42 
     43 - [x] Striping
     44 - [x] Mirroring
     45 - [ ] Parity
     46 
     47 RAID 10 (also referred to as RAID 1+0) combines mirroring and striping for reliability and performance. RAID 10 requires at least four disks and works by striping across pairs. This results in a 50% decrease in storage capacity w.r.t. RAID 0, but can survive one disk failure from each pair before any data is lost. Once data is lost, it is likely all data will be lost. While the worst case for this means it can only survive one failure, this is often not the case.