[ofa-general] Low NFS RDMA performance with Connect X

[Ramachandra K]

On 9/6/07, Talpey, Thomas <Thomas.Talpey at netapp.com> wrote:

> Can you post the full commandline of your NFS mount and iozone
> invocations?

iozone -Rab nfs_rdma_connectx_fw-2.2.xls -g 2G -c

mount.rnfs -o rdma=100.1.1.1 100.1.1.1:/home/rkuchimanchi/nfs_testdir tst

/etc/exports:

/home/rkuchimanchi/nfs_testdir 100.1.1.2(insecure,rw,async)

> I'm also curious if there were any NFS or RPC related
> messages appearing in the dmesg log during the run.

No, there were no special NFS or RPC related messages or errors during the run
except for the normal messages at startup and about connection establishment
and parameters.

>  Finally, were any RPC- or NFS-related patches applied to the RHEL5 kernel outside
> of the NFS/RDMA ones?

No. Its the normal RHEL 5 kernel with OFED-1.2.5 installed. I compiled
the NFS RDMA sources with OFED-1.2.5 sources and loaded the NFS RDMA
related modules.

Regards,
Ram