CyberGrind // Orange Book // Section 05

RAID — Redundant Array

of Independent Disks

Striping · Mirroring · Parity — trade-offs in performance, capacity, and fault tolerance

⚠ RAID IS NOT A BACKUP. It does not protect against accidental deletion, ransomware, or physical disasters. Always maintain separate, independent backups.
// The Three Techniques
⚡
Striping
Data is split into chunks and distributed across multiple drives. All drives read/write simultaneously — maximum throughput, zero redundancy.
🪞
Mirroring
Identical data is written to two or more drives simultaneously. If one fails, an exact copy remains. Simple and reliable.
🧮
Parity
Mathematical checksums of data blocks are stored across drives. If a drive fails, parity data allows the missing blocks to be reconstructed.
// RAID Levels
RAID 0
Striping — Performance Only
NO REDUNDANCY
Disk 1
A1
A3
A5
Disk 2
A2
A4
A6
Min Drives
2
Usable Space
100%
Drives Lost
0
Read Speed
★★★★★
Maximum read/write throughput
Full capacity — no overhead
One drive failure = total data loss
Video editing scratch disks, caching
RAID 1
Mirroring — Pure Redundancy
SAFE
Disk 1 (Primary)
A
B
C
Disk 2 (Mirror)
A
B
C
Min Drives
2
Usable Space
50%
Drives Lost
1
Read Speed
★★★☆☆
Simple; survives 1 drive failure
Fast read (can read from either disk)
50% of storage lost to mirroring
OS drives, small business file servers
RAID 5
Distributed Parity — Balanced
BALANCED
Disk 1
A1
B1
P
Disk 2
A2
P
C1
Disk 3
P
B2
C2
Min Drives
3
Usable Space
(N−1)/N
Drives Lost
1
Write Speed
★★★☆☆
Good capacity efficiency
Strong read performance
Parity slows writes; 1 drive tolerance only
General NAS, file servers, SMB storage
RAID 6
Dual Parity — Enhanced Safety
HIGH FAULT TOL.
Disk 1
A1
P1
P2
Disk 2
P1
B1
P2
Disk 3
P2
A2
B2
Disk 4
A3
P2
P1
Min Drives
4
Usable Space
(N−2)/N
Drives Lost
2
Write Speed
★★☆☆☆
Survives 2 simultaneous drive failures
Critical for large arrays with long rebuild times
Lowest write speed; needs 4+ drives
Large NAS, enterprise arrays, EMC/IBM
RAID 10
Mirrored Stripes — Best of Both
RECOMMENDED
Disk 1
A1
A3
Disk 2 (mirror)
A1
A3
Disk 3
A2
A4
Disk 4 (mirror)
A2
A4
Min Drives
4
Usable Space
50%
Drives Lost
Multiple*
Read/Write
★★★★★
Fastest performance with redundancy
Fast rebuild — mirror copy, no parity calc
50% capacity overhead (expensive)
Databases, VMs, mission-critical servers
// Quick Reference
Level Technique Min Drives Usable Space Drives Tolerated Speed Best Use
RAID 0 Striping 2 100% 0 Fastest Non-critical high-speed work
RAID 1 Mirroring 2 50% 1 Good reads OS drives, simple redundancy
RAID 5 Stripe + Parity 3 (N−1)/N 1 Good reads NAS, SMB file servers
RAID 6 Stripe + Dual Parity 4 (N−2)/N 2 Slower writes Large arrays, enterprise
RAID 10 Mirror + Stripe 4 50% Multiple* Very fast Databases, VMs, critical apps
RAID 50 RAID 5 + Stripe 6 Varies 1 per group Fast High-reliability high-perf
RAID 60 RAID 6 + Stripe 8 Varies 2 per group Good Large enterprise storage

* RAID 10 tolerates multiple failures as long as no two failed drives are in the same mirrored pair.

cybergrind.org // Orange Book // RAID Reference