RAID 5 Vs 6 on DMX

Question

Does anyone have real life experience or data they can share on RAID 5 vs 6 with a DMX array? I have spoken to representatives from EMC's SPEED group, but they did not provide much in terms of real life scenarios.

Based on the material I have been presented Raid 5 and 6 have similiar performance except for random writes and particularly small random writes. This type of IO exposes the the large stripe size and double parity writes of Raid 6.

Message was edited by:
Stefano Del Corno

removed sentence due to customer request

hermiecloete · Answer

I actually performed testing on a DMX4 with different drive types, 146GB, 300GB 400GB & 500GB using all RAID Types, R10, R5 (3+1), R5 (7+1), R6 (6+2) and R6 (14+2). Used dynamic cache provisioning to force spindle activity. In all the tests, R6 had about 40 to 50% lower performance than R1 and R5. Strongly suggested only for Archiving or Virtual tape.

xe2sdc · Answer

As you said, 'used DCP to force spindle activity' .. But your real-world workload usually won't reach WP limit... Thus probably you won't notice differences at all for writes And while reading, RAID5 or RAID6 may perform almost the same.. Did you run specific tests on read and write ??Message was edited by: Stefano Del Corno

RRR · Answer

EMC recommends R6 for the 1TB SATA drives, since rebuilding a hot spare can take many hours and if your RAID group is something like 7+1, it could even take a day. So what if a 2nd disk failure occurs ? That's why the R6 is there, to protect these multi TB raid groups.

hermiecloete · Answer

Yes I did, read/write. Diff block sizes, different work loads, random and sequential.

Symmetrix

RAID 5 Vs 6 on DMX

Was this post helpful?