7.6. RAID 10-Striped Mirror Sets

7.6. RAID 1+0 Striped Mirror Sets

The most expensive of RAID configurations, RAID 1+0 combines RAID 1 mirroring with RAID 0 striping. The drives are first mirrored and then striped across the member disks. RAID 1+0 has good performance, no write penalties, and excellent redundancy.

RAID 1+0 is illustrated in Figure 7-7.

Figure 7-7. RAID 1+0.

This array configuration can lose several disks as long as they do not belong to the same mirrored pair. If two disks in the same mirrored pair fail, the data is lost. This means that RAID 1+0 cannot guarantee protection against a two-disk failure.

RAID 1+0 process involves the following sequence:

1. Disk 0 is mirrored to disk 2.

2. Disk 1 is mirrored to disk 3.

3. Disk 0 and disk 1 are striped together.

4. Disk 2 and disk 3 are striped together.

In a RAID 1+0 configuration, all HP RAID controllers can

• Sustain multiple drive failures.

• Sustain an entire bus failure if the drives are equally distributed across buses.

• Service I/O requests to all operational drives in a degraded condition.

• Survive n/2 drive failures, where n is the number of drives in the array, as long as one member of each mirrored pair survives.

7.6.1 RAID 1+0 Performance

Performance characteristics of RAID 1+0 are the same as RAID 1. RAID 1+0 can tolerate multiple disk failures as long as no two failed disks are mirrored to each other.

7.6.1.1 READ PERFORMANCE

For read operations in a RAID 1+0, the disk controller can use only half of the disk drives because the other half contains the same information. This has a slight negative impact on performance compared to RAID 0. The number of disk drives (spindles) available for read request processing is reduced by 50%.

This negative impact on read performance is offset in the process of reading data from the drives; the requested data nearest to the read/write heads will be read.

7.6.1.2 WRITE PERFORMANCE

For write operations, the disk controller must generate the same write request twice. The same data must be written to both sets of mirrored disk drives causing the array controller to perform twice as many write operations, which generates a substantial amount of overhead.