Unsolved

This post is more than 5 years old

8 Posts

1197

June 15th, 2011 09:00

Celerra NS-960 10g throughput problem

Our customer have problems with linear read-write to celerra using cifs protocol.

Configuration:

6x15 2T ATA disks

3 FC loops

Created 8x14 RAID5 groups

In celerra all this raid groups populated into default pool.

From this pool we create 1x16TB fs (celerra limit) and presented to cifs server.

We also create lagg using 2x10g ethernet devices, then connect it to cisco nexus.

For test we used HP DL380G7 with 2 port 10g adapter with only 1 active connection. For temp storage we used same Clariion with 14x400gb raid5 FC disks on another backend loop and 8g FC card.

We try to put 300gb file to celerra and then read it...

Average read speed was 180-220mb/s

Write 160-200mb/s

Celerra reported about 15-20% cpu usage.

System has 8g link to AUX. There is speed? How we can find whats the problem? We need expected 600-800+ mb/s in linear r/w... Or 220mb/s is normal for NS-960?

On next week we plan to test same config using NFS...

11 Legend

 • 

20.4K Posts

 • 

87.4K Points

June 15th, 2011 11:00

it would be interesting to look at server_stat output for dvols ..to see how many dvols are utilized concurrently.  Maybe manual volume management would provide better striping than native AVM and give your better throughput.

31 Posts

June 27th, 2011 12:00

I am having a similar problem with horrible CIFS perfomance issues.  We have a VG2 that is connected to a Nexus with two 10G links in an LACP trunk.  The VG2 is direct attached to our CX4-480.  When copying to a CIFS share I'm lucky if I average 3 - 5MB/s.  I am at dart level 6.0.40-5 and flare version 4.30.000.5.509.  I have an SR open and they want to try upgrading the flare to 30/517.  I'm not sure that's goign to solve the issue; but I know 517 does solve a memory leak issue.  Anyone else out there having CIFS performance issues?

No Events found!

Top