FibreMAX support???

[Bob Zelin]

ouch ! Isn’t this a 16 drive array. You should be getting triple that performance. ProMax didn’t make this array, they bought it from a company in the US, so their support is here. I will try to find out the companies name.

Bob Zelin

[Ian Karr]

Thanks for your help, Bob.

Yes, it’s a 16 drive unit…but only 10 drives currently in it. 5 on each side of the array. (configured left-to-right as drives in the first five slots, then three empty, then 5 drives, then 3 empty).

In the admin program two RAIDS show up…RAID 0 and RAID 1. I’ve sliced each one into 5 equal slices. I’m trying to use disk utility to software stripe the slices from each RAID together end up with 5 x RAID 50 “partitions”. No matter what I do-stripe or not, the performance is exactly the same.

Bob, if you have a moment, please email me (ian@ikacollective.com). I may have an opportunity for you…

Thanks!

Ian

[Sean Oneil]

You probably have a bad disk in there.

Sean

[Ian Karr]

Sean,

Thanks for the feedback, but it appears not. I’ve also tried striping all 10 disks together as one unit, and still no performance difference. It’s always around 180mb/s.

I’m beginning to think it might be related to the fibrechannel interface. Both ports light up and seem to be okay, but If only one port is working, that might explain the limitation.

Mr. Zelin???

[Sean Oneil]

You misunderstood. I’m saying that one of your 10 disks is bad. Stripping all 10 together when one of them is choking is going to cause the exact problem you have described. Creating two groups of 5-each and stripping those together is no different. If one failing drive is in there, the whole thing will only perform as fast as the failing one can operate.

Fibre channel ports don’t go bad. Hard drives do. I can almost grantee this is your problem. Create two separate groups (don’t stripe them together) and test them individually. If one is fast while the other is slow, then you definitely have a bad disk in there.

Sean

[Ian Karr]

Thanks so much for your help and feedback, Sean.

Unfortunately, I already did that test. When the drives were two Raid sets (of 5), not striped together, both sets had the same throughput. I suppose, according to your theory, one drive on each side could be bad.

Also, all 10 drives show up with green lights and seem to check out. I suppose the next test I’ll do is unbundle the raid and check the throughput on each drive by itself.

[Bob Zelin]

this is the correct test. I realize how time consuming this is. Each drive should give you about 68mb/sec. Anything showing 40mb/sec or lower is a bad drive.

Bob Zelin

[Simon Blackledge]

Can’t you just set the raid to JBOD ?

[Ernesto Sanchez]

The two companies that can help are either GMAX or AC&NC. The former is ProMax’s storage division that spun off and the latter distributes the box FibreMax is based on.

https://gmaxstorage.com/
https://www.acnc.com/index.html

Cheers,
Ernesto (Formerly of ProMax)

Cheers,
Ernesto Sanchez
Senior SAN Engineer
THE DR GROUP
O – 323.896.3416
C – 949.690.3680
ernesto@thedrgroup.com

[Ian Karr]

Bob, Ernesto, Everyone…

Thanks so much for the help.

Well, it took a while, but I found out what was up. It wasn’t the drives at all. It turns out there was a bottleneck getting through the fibrechannel interface. Specifically, the arrays and partitions were all configured to travel through FC port 0. FC port 1 had no traffic. Basically the throughput of the RAID exceeded the bandwidth of the single channel. That’s why, no matter how many drives I striped together, the speed remained the same.

So…

I reconfigured the box. Set up 2 RAID arrays, then partitioned them into 5 separate LUNs. The volumes on RAID 0 were directed to FC port 0, RAID 1 was directed to FC port 1. Then, in disk utility, I striped equal-sized partitions from both RAIDS together to create Raid 50s. The thing is bookin’ now.

Ian