Storage & RAID

PC & IT Hardware Servers / Guide

Storage & RAID

HDD vs SSD, RAID levels, and when to use each ߀” explained simply.

Storage Technologies

HDD ߀” Hard Disk Drive

Mechanical spinning platters. Data is read/written by a moving head.

  • Speed: 100߀“200 MB/s sequential; high seek latency (~10 ms)
  • Capacity: Up to 24 TB (consumer), 32 TB (enterprise)
  • Durability: Sensitive to shocks; not suitable for mobile use
  • Cost: Cheapest per GB ߀” best for bulk/cold storage

SATA SSD ߀” Solid State Drive

Flash memory with a SATA interface. Limited by the SATA III interface cap.

  • Speed: ~550 MB/s sequential read; ~520 MB/s write
  • Latency: ~0.1 ms ߀” massively faster than HDD for random I/O
  • Form factors: 2.5" (most common), M.2 SATA

NVMe SSD ߀” Non-Volatile Memory Express

Flash memory connected directly to the PCIe bus, bypassing SATA overhead.

  • Speed: 3,000߀“7,000 MB/s sequential (Gen4/Gen5); far higher IOPS than SATA SSD
  • Form factors: M.2 (consumer/workstation), U.2/U.3 (enterprise), PCIe add-in card

SAS ߀” Serial Attached SCSI

Enterprise interface used in servers. Faster protocol than SATA, supports dual-porting.

  • More reliable at sustained 24/7 workloads
  • Supports SAS HDDs (15,000 RPM), SAS SSDs
  • Requires a SAS HBA or RAID controller

RAID ߀” Redundant Array of Independent Disks

RAID combines multiple drives to provide redundancy, performance, or both. RAID is not a backup ߀” it protects against drive failure but not accidental deletion, ransomware, or multiple simultaneous failures.

RAID Levels

RAID 0 ߀” Striping

Data is split across all drives. No redundancy.

  • Benefit: Maximum performance and capacity
  • Risk: One drive fails ߆’ all data lost
  • Minimum drives: 2
  • Use case: Temporary scratch space, video editing cache

RAID 1 ߀” Mirroring

Data is written identically to two drives.

  • Benefit: Full redundancy; can lose one drive and keep running
  • Usable capacity: 50% (2 G— 4 TB = 4 TB usable)
  • Minimum drives: 2
  • Use case: OS drives, small critical datasets

RAID 5 ߀” Striping with Parity

Data and parity information are distributed across all drives.

  • Benefit: Can lose one drive; good read performance
  • Usable capacity: (N-1) drives (3 G— 4 TB = 8 TB usable)
  • Minimum drives: 3
  • Use case: General purpose file servers

RAID 6 ߀” Striping with Double Parity

Like RAID 5 but with two parity blocks.

  • Benefit: Can lose two drives simultaneously
  • Usable capacity: (N-2) drives (4 G— 4 TB = 8 TB usable)
  • Minimum drives: 4
  • Use case: Large arrays where rebuild time is long and double failure is a risk

RAID 10 ߀” Mirrored Stripes (1+0)

Combines RAID 1 and RAID 0: data is mirrored, then striped.

  • Benefit: High performance and redundancy; fast rebuilds
  • Usable capacity: 50%
  • Minimum drives: 4
  • Use case: Databases, high-I/O applications

RAID Controllers

Type Description
Software RAID (OS) CPU handles parity; free, flexible, slightly slower
Fake RAID (motherboard) Marketed as hardware RAID; actually software. Avoid for critical data
Hardware RAID card Dedicated processor and cache (BBU); best performance and reliability

Popular hardware RAID controllers: LSI MegaRAID, Broadcom, Adaptec.

Rebuild Time Warning

When a drive fails and is replaced, the array rebuilds by re-computing parity across all remaining drives. During rebuild:

  • Array is vulnerable to a second failure
  • Performance degrades significantly
  • Larger drives = longer rebuild (a 16 TB drive can take 24߀“72 hours)

This is why RAID 6 (double parity) is recommended for large arrays.

Back to Servers