A900 performance

[Jackiez]

Hi, ONTAP experts,

        We have a A900 performance case.  We use fusion to get 33GB/s + performance data(32KB block, 100% sequential read). But at customer site, we test the performance , only get 23GB/s(32KB block, 100% sequential read, 2 Aggr per controller, total 4 Aggr, 16 volumes) ,latency 5ms. we use 9 client (100Gb NIC per client) to connect one 100Gb switch , then connect 8 ports to A900 , ONTAP version is 9.12.1P3. 

    Would you share some experience on how to optimize A900 performance (especially Bandwidth) ?

    (BTW , when we use 1MB block,  get 30GB/s , but just keeps a few seconds, not consistent)

[TMACMD]

You are lacking so much information for the community to be helpful....

What is/are the client(s)? how are they connected? NFS/CIFS? v3/v4/v4.1/v4.2? SMB2/SMB3?

Operating systems? mount parameters...

I mean...give us something

[Jackiez]

Hi, TMACMD,

        Thank you !  Here is more details information : client ,ubuntu 18.04, nfs3.0; mount parameters :

mount -t nfs -o vers=3,nolock,proto=tcp,rsize=1048576,wsize=1048576,hard,10.199.199.105:/vol1 /netapp01

[TMACMD]

If you haven’t done so, you should really take a hard look at https://www.netapp.com/media/10720-tr-4067.pdf

there are many details in this doc that should assist with improving performance. 

second, you are using a large read/wire size. If I recall, ONTAP defaults to 64k and you are teeing to use 1M. You will need to modify the nfs svm then remount the clients

[Jackiez]

Hi, TMACMD,

        Thanks for your advice !

        We  set  some parameters ,   tcp slot table =128,  nfs tcp -max-xfer-size=1M ,  then we got much better performance data:  32K block,  28-30GBps;  1MB block, 34-36GBps.    Match the sizing performance data.