NFS on 10G interfaces still painfully slow

[Rick Macklem]

Alan Somers wrote:

>On Tue, Aug 2, 2016 at 2:49 AM, Gerrit Kühn <gerrit.kuehn at aei.mpg.de> wrote:
>> Hi all,
>>
>> I already reported this issue here a year ago and unfortunately was not
>> able to fix it back then. Now I had another run at it, using two recent
>> 10.3-machines with a direct 10G link. I still see nfs is painfully
>> slow (around 20-80MB/s). I tried both nfsv3 and nfsv4, with almost the same
>> results. Everything I tried so far (mtu size, wcommitsize, readahead...)
>> only makes things worse or at least not much better.
>> Moving data in different ways (scp, ggate) is much faster, so plain
>> network speed should not be an issue.
>>
>> Is there anyone around here who can confirm that nfs can go faster over
>> 10G links?
>> Any hints for further tuning/debugging are greatly appreciated.
>>
I can't help much, but a couple of things you can try:
- Disable TSO
- Turn off/reduce interrupt moderation on the net interface. (NFS perf.
  depends on response time and anything that delays interrupt servicing
  will slow it down.)

Good luck with it, rick

>>
>> cu
>>   Gerrit
>
>
>I can get 1GB/s over NFS on a 10G link, so it's not always slow.
>There's probably something about your setup that's slowing it down.
<What is your NFS client?  If Linux, make sure that you're using the
>"async" mount option instead of "sync".  What benchmark are you using
<to measure that speed?  Did you remember to start lockd and statd?  If
>you post your /etc/exports and the client's /etc/fstab, that might
>reveal something.
>_______________________________________________
>freebsd-net at freebsd.org mailing list
>https://lists.freebsd.org/mailman/listinfo/freebsd-net<https://lists.freebsd.org/mailman/listinfo/freebsd-net>
>To unsubscribe, send any mail to "freebsd-net-unsubscribe at freebsd.org"