RAID-6 and double-parity performance

In this lesson, you will learn how RAID-6 provides robust data protection through double parity and why this comes with a specific performance cost during write operations.

Understanding Double Parity

While RAID-5 uses a single parity block to recover from a single drive failure, RAID-6 uses two independent parity blocks (often labeled P and Q). This allows the array to survive the simultaneous failure of any two drives. These parity blocks are distributed across all drives in the array, ensuring no single disk becomes a performance bottleneck for parity calculations.

The Write Penalty

The "cost" of this reliability is known as the write penalty. Every time you write data to a RAID-6 array, the system must perform more calculations compared to RAID-5. Because there are two parity blocks, a small write requires:

1. Reading the existing data.
2. Reading the existing P and Q parity blocks.
3. Calculating the new P and Q values based on the changes.
4. Writing the new data, the new P parity, and the new Q parity back to the disks.

This is often referred to as the "6-write penalty" (as opposed to 4 in RAID-5). Because of this overhead, you might notice slower write performance, especially if your hardware controller or CPU is busy.

Observing Performance

You can monitor how your RAID array is handling these operations using iostat. When performing heavy write tasks on your RAID-6 volume, look for high %util (utilization) and await (average wait time) values:

# Monitor your RAID array performance
iostat -dx 1

If you notice that your write speeds are significantly lower than your read speeds, it is usually because the system is busy recalculating those two parity blocks for every single write command. While RAID-6 is excellent for high-reliability storage where data loss is unacceptable, it is best paired with LVM caching strategies (which we covered in previous lessons) to help absorb these write bursts and hide the latency caused by double-parity calculations.

Key Takeaway: RAID-6 provides superior fault tolerance by allowing for two concurrent drive failures, but it introduces a significant write penalty due to the dual-parity calculation requirement for every write operation.