http://bugzilla.novell.com/show_bug.cgi?id=557760 http://bugzilla.novell.com/show_bug.cgi?id=557760#c10 Jon Nelson <jnelson-suse@jamponi.net> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |ASSIGNED Info Provider|jnelson-suse@jamponi.net | --- Comment #10 from Jon Nelson <jnelson-suse@jamponi.net> 2009-12-15 02:26:59 UTC --- I would ignore the read speeds for now. The write speeds are pretty bad. Can NFSv4 work over UDP? I just re-tested. On an otherwise totally idle server and client, with no other meaningful traffic on the switch whatsoever: 101911552 bytes (102 MB) copied, 59.5694 s, 1.7 MB/s The server's storage looks like this: 4x SATA disks arranged as a raid6, lvm on top of that, and jfs (for this filesystem) on top of that. Local I/O for comparison (caches dropped): 33.2MB/s. Not awesome but substantially better than 1.7MB/s. I gave UDP a try. It did not go well. I got a *lot* of this: [ 2535.582524] nfs: RPC call returned error 88 [ 2535.582543] nfs: RPC call returned error 88 [ 2535.582554] nfs: RPC call returned error 88 [ 2535.582563] nfs: RPC call returned error 88 slightly less than 3000 lines worth, AND, the following gem: jnelson@frank:~> dd if=/dev/zero of=/multimedia/foo bs=1k count=100000 dd: closing output file `/multimedia/foo': Socket operation on non-socket jnelson@frank:~> Whaaa? and then lots and lots of I/O errors. So I gave up on UDP (NFSv4). How large of a tcpdump would you need? My last (quick) test was: 101963776 bytes (102 MB) copied, 103.186 s, 988 kB/s Mount options: rw,noatime,nodiratime,hard,intr,noacl,rsize=32768,wsize=32768 Raw TCP sends from this box: 111303.89kB/s -- 109MB/s for blasting TCP from the client to the server as fast as they can each send/receive. I don't think it's the hardware. ;-) (as a side note, setting the MTU on the client to 8200 makes performance DROP considerably: to a mere 72MB/s) I'm at your disposal. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.