FAS 3140 performance issue?

[sdrcpdpune]

Hi,

I have 2 nodes of FAS3140 in a cluster. Since 1 month I am facing a strange performance issue in it.

1. Users are reporting that tasks (either copying from filer or starting an installation) from filer are now taking 3-10 times more time to do. This was specially observed in peak hrs.

2. I checked with the latency of volumes and those volumes have less than 5 latency at that time.

3. I also checked the bandwidth utilization of filer ports as well as at cisco switches which are available between users and filers, and they are only 20-30% utilized.

4. What will be the max latency and number of IOPS that 3140 can handle?

Can someone help me in identifying the problem?

Thanks.

[brenetapp]

latency - try this

from console type:

stats show -i 10 volume:VOLume_NAME:read_latency volume:VOLUME_NAME:Write_latency

(VOLUME NAME - just change name )

-i 10 - time periot between test in sec.

results:

Read_latency / Write_latency  should be less then 2000 for fileservers

0-1000		Good
1000	2000	Reasonable
2000	5000	Busy
5000		Bad

IOPS = ~150-200 IO/per disk in your AGGR if disk are FC,,,, 100-120/per disk

[sdrcpdpune]

Hi,

Output of that command is:

> stats show -i 10 volume:vol3:read_latency volume:vol3:write_latency
Instance read_latency write_latenc
                   us           us
vol3     31617.50            0
    vol3            0            0
    vol3            0            0
    vol3     11179.10            0
    vol3      8183.47            0
    vol3     16807.36            0
    vol3     11206.60            0
    vol3     12059.46            0
    vol3      8381.23            0
    vol3      4848.33            0
    vol3      6196.23            0
    vol3      6456.89            0
    vol3      8402.19            0
    vol3            0            0
    vol3            0            0

What is the solution now?

[radek_kubka]

Hi,

How much space is left in your volume and the containing aggregate? If it's tiny (say less than 10% free space), then poor read performance may be the outcome of fragmentation.

If that's the case, then the first thing to do would be to provide more 'breathing space' for your volume.

Here are tons of information re fragmentation & reallocation:

http://communities.netapp.com/thread/6530

Regards,

Radek

[brenetapp]

appears to be quite busy

how many drives you have in aggregate? do you have any other volumes on this aggregate?

type command: stats show -i 10 volume:vol3:read_ops volume:vol3:write_ops volume:vol3:total_ops

results of command divide by number of disk,

* intable should be one zero more (1000us -- 1ms )

[mitchells]

Do you take a lot of snapshots on vol3?

Are you using this volume as a snapmirror source?