Skip to content

Latest commit

 

History

History
145 lines (111 loc) · 3.62 KB

11-RAID.md

File metadata and controls

145 lines (111 loc) · 3.62 KB
Raw
title subtitle author date
11.1. RAID
Linux<br/>HOGENT toegepaste informatica
Thomas Parmentier, Andy Van Maele, Bert Van Vreckem
2024-2025

RAID - Redundant Array of Independent Disks

JBOD - Just a Bunch of Disks

JBOD

RAID 0 - Striping

RAID 0

  • Secure against single theft
  • Unsecure against hardware failure

RAID 1 - Mirroring

RAID 1

  • Safe against HW failure
  • COMPLETELY unsafe against theft
    • Use volume encryption! LUKS, Bitlocker, ...
  • Hot spare disk = idle disk, if needed a rebuild action can start automatically

Parity calculations: XOR

A B XOR
0 0 0
0 1 1
1 0 1
1 1 0

Calculating parity

data1:  0100 1110
data2:  1010 1011
-----------------
parity: 1110 0101

Recovering data1 using parity:

data2:  1010 1011
parity: 1110 0101
-----------------
data1:  0100 1110

RAID 5 - Striping with parity

RAID 5

  • At least 3 disks needed
    • One disk lost for capacity
  • Safe against HW failure of one disk
    • One disk fails ⇒ RAID 0-like situation
    • If hot spare is configured, rebuild action starts
    • Warning: increased load on other disks, may trigger other failures
  • Current processors make RAID 5 as fast as RAID 1 (or better!)

Parity with 4 disks

block1a:    0100 1110
block2a:    1010 1011
---------------------
parity-tmp: 1110 0101
block3a:    1101 0011
---------------------
parity:     0011 0110

Recovering block2a using parity

block3a:    1101 0011
parity:     0011 0110
---------------------
parity-tmp: 1110 0101
block1a:    0100 1110
---------------------
block2a:    1010 1011

RAID 10 - Mirroring + striping

RAID 10

  • Is RAID 1 (mirroring) combined with RAID 0 (striping)
  • At least 4 disks needed
    • Half capacity usable!
  • Very frequently used in enterprise environments

Software vs Hardware RAID

  • Difference = where are the calculations done?
    • HW RAID: dedicated RAID chip (controller - onboard or plugin card)
    • SW RAID: CPU
  • Directly Attached Storage (DAS): not much use for HW RAID
    • Main CPU is strong enough for added RAID functionality
    • Recovery/forensic analysis is easier with SW RAID
  • Network Attached Storage (NAS)/Storage Area Network (SAN): HW RAID
    • Not much choice, often embedded in the device!
    • Dedicated RAID controller with proprietary firmware

The big problem with HW RAID

  • Proprietary firmware
    • In case of data loss: recovery/forensic analysis is difficult/impossible
  • Good RAID controller duplicates config on each disk for recovery purposes
    • Not a standard! Depends on implementation
  • RAID controllers are aging
    • E.g. RAID 10 on Areca RAID controller is not recognized if the disks are connected to an Adaptec RAID controller
    • OR: RAID 10 on a 5 year old Areca RAID is not recognized in recend Areca RAID controller
    • RAID controller has become a single point of failure!

Software RAID

  • Can be done by the OS
    • Linux Software RAID (mdadm), Linux Volume Management (LVM)
    • Windows Dynamic Volumes, Storage Spaces
    • 3rd party software (e.g. Starwind)
  • Usually easily expandable
  • Recoverable
  • Backwards compatible
  • "Free" (included in OS)
  • Independent of interface (IDE, SATA, SCSI, ...)
  • Fast (with modern CPUs)

Disadvantages of SW RAID

  • CPU load
  • OS is not easy to install on a RAID volume
  • Not all RAID levels are supported
  • Hot swap not always supported