How RAID Controllers Ensure Data Integrity and Performance: A Deep Dive

Disk fault tolerance ensures data integrity and processing capability when a subsystem experiences hard‑disk errors or failures. RAID controller cards achieve this on RAID 1, 5, 6, 10, 50, and 60 through redundant disk groups.

1 Consistency Check

For RAID levels with redundancy (1, 5, 6, 10, 50, 60), the controller can verify data consistency across disks, compare with redundant data, and automatically repair inconsistencies while logging errors. RAID 0 lacks redundancy and therefore does not support consistency checks.

2 Hot Spare

The hot‑spare feature is provided by hot spare disks and emergency backup.

Hot Spare Disk

A hot spare is an idle disk that automatically replaces a failed member disk and reconstructs its data.

In the management interface or CLI, an idle disk with equal or greater capacity and the same media type and interface as the member disks can be designated as a hot spare.

Supported hot spares:

Global hot spare: shared by all configured RAID groups on the controller; multiple global spares can be configured.

Local hot spare: dedicated to a specific RAID group; each group can have one or more local spares.

Hot spares are only used for RAID groups with redundancy (RAID 1, 5, 6, 10, 50, 60) and replace a failed disk on the same controller.

Emergency Backup

If a RAID group with redundancy experiences a disk failure and no hot spare is assigned, an idle disk will automatically replace the failed member and reconstruct data, preventing data loss. The replacement disk must have capacity equal to or greater than the member disk and the same media type.

3 RAID Reconstruction

When a disk fails, the controller can reconstruct the data onto a new disk. Reconstruction applies only to RAID levels with redundancy (1, 5, 6, 10, 50, 60).

If a hot spare is available, it automatically replaces the failed disk and starts reconstruction. Without a hot spare, reconstruction can only begin after a new disk is installed. If the system powers down during reconstruction, the controller resumes the task after restart.

The reconstruction rate (CPU usage) can be set from 0% (run only when the system is idle) to 100% (use all CPU resources); users should choose an appropriate value based on workload.

4 Virtual Disk Read/Write Policies

When creating a virtual disk, a read/write strategy must be defined.

Read Policy

The controller supports two read strategies:

Read‑ahead (e.g., "Always Read Ahead", "Read Ahead", "Ahead"): data is prefetched into cache, reducing seek time and improving read speed. This requires power‑loss protection; a faulty super‑capacitor may cause data loss.

Non‑read‑ahead: data is read only when a read command is received, without prefetching.

Write Policy

Write‑Back: data is first written to cache and later flushed to the virtual disk, improving write speed. Requires power‑loss protection.

Write‑Through (direct write): data is written directly to the virtual disk without caching; does not require power‑loss protection but is slower.

Write‑Back with BBU: when a Battery Backup Unit is present and healthy, writes go through cache; if the BBU is absent or faulty, the controller automatically switches to Write‑Through.

Write‑Back Enforce: forces Write‑Back even if the controller lacks a capacitor; not recommended because data may be lost on unexpected power loss.

5 Data Power‑Loss Protection

Principle

Data is written to the controller's cache faster than to the disks, improving performance but risking data loss on sudden power loss.

Enabling cache boosts write performance; when the cache fills, data is flushed to disks.

However, a power outage can cause loss of cached data.

Super‑capacitors can be added to preserve cache data during power loss by writing it to NAND flash.

Super‑Capacitor Calibration

The controller performs a three‑stage charge‑discharge cycle to keep the capacitor voltage stable, automatically switching the write policy to Write‑Through during calibration.

6 Disk Striping

Striping balances I/O load across multiple physical disks, dividing data into small blocks stored on different drives, enabling parallel access and improving performance, though it does not provide redundancy.

Key concepts:

Stripe width: number of disks participating in the stripe.

RAID group stripe size: size of data blocks written simultaneously to all disks in the group.

Disk stripe size: size of the block written to each individual disk.

7 Disk Mirroring

Mirroring (used in RAID 1 and RAID 10) writes identical data to two disks, achieving 100% redundancy; if one disk fails, data remains available without interruption.

8 External Configuration

External configuration refers to RAID metadata present on a newly installed disk or after a controller replacement. The controller can import, delete, or ignore such configurations based on the server's needs.

9 Disk Power‑Saving

The controller can spin down idle SAS/SATA disks to save energy. Disks wake up when operations such as RAID creation, hot‑spare addition, or dynamic expansion occur.

10 Disk Pass‑Through (JBOD)

Enabling JBOD allows direct command pass‑through to connected disks without RAID processing, useful for OS installations or applications that need raw disk access.