RAID-5 Performance with SoftRAID 5

RAID-5 offers fault tolerance against a drive failure. Performance at least on reads should be equivalent to a RAID-0 stripe with one less drive, e.g., the performance of a 4-drive RAID-5 should be similar to a 3-drive RAID-0 stripe.

See also SoftRAID RAID-4 / RAID-5 Performance with Heavy CPU load.

Tested with very fast hard drives.

Performance of RAID 5 with 4 or 5 drives

Performance is outstanding, faster than many hardware RAID-5 solutions.

With one OWC Thunderbay, a 4-drive array can be created. Addition of a 5th drive is possible using another Thunderbay, which raises the read performance, but does not improve the write performance. That behavior might be improved with the release version.

Tested with DiskTester on the late 2013 MacBook Pro in dual OWC Thunderbay enclosures, each on its own Thunderbolt port. SoftRAID 5 beta.

disktester run-area-test --iterations 5 --test-size 4G --delta-percent 5

RAID 5 performance with 4 and 5 hard drives, SoftRAID 5 beta

Performance with transfer size

Performance varies by transfer size; smaller transfer sizes start losing any multi-drive benefits (stripe size).

Transfer sizes exceeding 32MB are likely unrealistic for real applications.

disktester run-sequential-suite --iterations 3 --starting-xfer 128K --test-size 4G --xfer-steps 4

RAID 5 performance with 4 hard drives vs transfer size, SoftRAID 5 beta

Performance of RAID 5 with 8 drives

Using dual OWC Thunderbay units and two Thunderbolt ports.

Performance is outstanding for reads. Write performance is gated to ~500MB/sec, but that might change with the release version. Consistency for both is excellent.

Tested with DiskTester on the late 2013 MacBook Pro in dual OWC Thunderbay enclosures, each on its own Thunderbolt port. SoftRAID 5 beta.

disktester fill-volume --xfer 4M

RAID 5 performance with 8 hard drives, SoftRAID 5 beta