[Cialug] Why is rsync/SSH so much faster than rsync/nfs?

[Matthew Steven]

Hello,

I'm working on setting up a raid 10 nas box (lucid), and would like to 
tune nfs to perform well enough to at least approach the throughput of 
the array so I can effectively run diskless workstations.

In moving the data to this new box however, I found using rsync over nfs 
is very slow indeed. I've played with a number of settings, particularly 
rsize/wsize, currently this is what the client /proc/mounts reports:

rw,noatime,vers=3,rsize=1048576,wsize=1048576,namlen=255,hard,
proto=tcp,timeo=600,retrans=2,sec=sys,mountaddr=192.168.0.15,
mountvers=3,mountproto=tcp,addr=192.168.0.15

iostat on the server shows that it's loafing while rsync is copying over 
nfs.

When I switch to using rsync over ssh to copy, the transfer speeds, 
particularly with lots of smaller files, goes way up. I don't know if 
it's because rsync is filling the packets more efficiently when 
operating over ssh, or if nfs is badly configured?

Seems to me that nfs should have a bit less overhead than ssh.

Cheers,
Matt

-- 
Matthew Steven
http://www.geniusweb.com/
(515) 999 0842