RAID 50 loses half the expected speed?

[Alex Gerulaitis]

Any chance the R380 is in a 4x slot? I believe it’s PCIe 1.0 and the 4x config could partially explain the 800MB/s speed cap… I think…

— Alex Gerulaitis | Systems Engineer | DV411 – Los Angeles, CA

[Blase Theodore]

Nope its in slot2 on a MacPro 5,1. Which is an x16 slot. And definitely an x8 connection, as verified by the connection speed in the ATTO util and the osx system info.

[Alex Gerulaitis]

I’ve never tried anything similar on a R380 so all I can offer are wild guesses.

Is it possible to reformat the array in RAID0 across all drives and test speeds? (It’d be a PITA to initialize RAID5 again – yet this would be a good test to measure R380 max throughput.)

What does ATTO say about R380 performance ceiling in RAID5?

Also, R380 seems to support RAID50 natively. Any reason you decided on soft (OS) striping?

[Blase Theodore]

Thanks, I should have mentioned that as well…
I did a test with RAID0’s instead of RAID5’s, but the results were the same.

The spec sheet shows native RAID50 support, but they don’t actually support it.

[Alex Gerulaitis]

It looks like R380 flatlines at 800MB/s. Puzzling though that RAID50 gives you worse performance vs. RAID5 with half the drives. Possibly it slows down as it works with more drives / RAID groups.

R680 then, or perhaps ARC-1882x.

— Alex Gerulaitis | Systems Engineer | DV411 – Los Angeles, CA

[Jon Schilling]

Sounds like a issue to pose to ATTO. R380 = 3G speed. R680 = 6G speed. Alex is right, I’d try an R680 card, or Areca 1882X card.

Jonathan Schilling
Vertical Sales Manager
Proavio Storage by Enhance Technology Inc.
12221 Florence Ave.
Santa Fe Springs, CA 90670
Dir: 562-777-3498
Main: 562-777-3488 X106
Fax: 562-777-3499
Email: jon@proavio.com

[Blase Theodore]

To be clear, we’re all on the same page of the specific problem I’m asking about?

OSX software striping fails.
700Mb/s + 700Mb/s (striped) = 500Mb/s.
It should equal 1200-1400.

I have 2 separate R380 cards. Each with its own RAID5, getting 700Mb/s.

I stripe them together through OSX and I get 500Mb/s. The ATTO cards are clearly working and doing their job correctly. The failure is in the last step.

[Alex Gerulaitis]

[Blase Theodore] “I have 2 separate R380 cards”

Not what you said initially:

[Blase Theodore] “each xtore feeds a single SAS connection to the R380
OSX 10.8.5

RAID SETUP:
12 drives > xtore_unit1 > SAS1 > RAID5 > “RaidGroup1”
12 drives > xtore_unit2 > SAS2 > RAID5 > “RaidGroup2″”

The above implies and describes a single R380.

You also mentioned that RAID0 exhibits the same behavior, which points to the card as the culprit.

If you actually need help figuring out why striping two RAID5 groups across two R380 cards causes slower performance, perhaps you’d want to list your configuration more precisely?

[Blase Theodore]

Sorry Alex, I thought I had posted a follow up in which I repeated the setup, but split across 2 cards, and the result was the same.

I must have hit “preview post” but then never actually posted. I’ll retype it below…

ALTERNATE TEST SETUP:

HARDWARE SETUP:
24 WD RE3 enterprise SATA 3Gb/s drives
dual ATTO R380 cards (3Gb SAS)
dual xtore 12bay JBOD expander units (3Gb SAS)
each xtore feeds a single SAS connection to a separate R380
OSX 10.8.5

RAID SETUP:
12 drives > xtore_unit1 > R380-1 > RAID5 > “RaidGroup1”
12 drives > xtore_unit2 > R380-2 > RAID5 > “RaidGroup2”
Interleave=512 / Sector size=4kb / Speedread=always / prefetch=0
Disk utility > “RaidGroup1” + “RaidGroup2”

In the alternate test setup each RAIDgroup is about 700Mb/s R/W. However once again, when striped together by disk utility, the resulting speed is only 5-600Mb/s. I would expect 12-1400Mb/s.

And to clarify the RAID5 vs RAID0..

I repeated both hardware test setups using RAID0’s instead of RAID5’s. The RAID0’s performed exactly as the RAID5’s did, and the overall results of all 4 tests were identical. Everything was always successful until the last step of striping them together.

[Bob Zelin]

let be me even more clear, so there is no mis-interpretation of what is going on here.

Blasé has TWO ATTO R380 cards. He has TWO AIC XStor 3gig chassis with 3gig drives. So this is not new equipment, that is capable of doing the usual 1500 MB/sec that we are used to seeing. He wants to COMBINE the two chassis in a RAID 50, for the SOLE PURPOSE of increasing his speed, so he does not have to go out and purchase new host cards, new 6g chassis, and new 6g drives. The drives on their own work perfectly fine. But when he merges the two together to create a single RAID 50, the speed DOES NOT INCREASE. This is his only problem.

You could make the same argument about this with all new equipment, (and an ATTO R680 or Areca 1882x, using TWO of these new cards, and two NEW 6g chassis with all 6g drives). The idea is – no matter how fast your single drive array is – if you merge both of them as a RAID 50 (or 60), shouldn’t the speed INCREASE when it appears as a single volume ?

Bob Zelin

Bob Zelin
Rescue 1, Inc.
maxavid@cfl.rr.com