Slow NFS Writes on F800 nodes

[Anubhav Ahuja]

Hi All, 

We are doing some SAS testing on a 4 node F800 cluster. We have a one to one export to VM mapping. Each VM is a 8 CPU 100GB memory machine reading/writing to one export. Total 8 VM's. We used the below mount options to mount the export on the VM. 

vers=3,acdirmin=0,noatime,nodiratime,rdirplus

vers=4,acdirmin=0,noatime,nodiratime,rdirplus

rw,vers=4,tcp,hard,intr,retry=2,retrans=5,timeo=600

rw,noatime,nodiratime,vers=3,rsize=131072,wsize=524288,namlen=255,acdirmin=0,hard,proto=tcp,timeo=600,retrans=5,sec=sys

rw,vers=4,tcp,hard,intr,retry=2,retrans=5,timeo=600,rsize=1048576,wsize=1048576

rw,vers=4,tcp,hard,intr,retry=2,retrans=5,acdirmin=0,noatime,nodiratime,rdirplus,rsize=1048576,wsize=1048576

But we are not getting the desired performace. The max reads and writes we get are 

reads 8339 mb/sec

writes 5043 mb/sec

Do you recommend any other mount options or tweaks in Isilon to improve the performance?

Thanks in Advance 

Anubhav

[Jean-Didier stefaniak]

You really want to share your testing methodology/topology here along with your realistic expectations.
otherwise anyone here can just give you a shot-in-the-dark suggestion; to start with : https://www.dellemc.com/en-us/collaterals/unauth/white-papers/products/storage/h16857-wp-onefs-best-practices.pdf

[John Beranek - PA]

Also:  https://www.dellemc.com/resources/en-us/asset/white-papers/products/storage/h17240_wp_isilon_onefs_nfs_design_considerations_bp.pdf 

John