Hi.
I've been needing to use amd lately to mount remote NFS exported fs.
AFAIK, the automounted support only NFSv2 which would not be an issue if
it was not so slow... reading is fine but writing is like 6 to 7 times
slower than with NFSv3.
The tests were made with a current/macppc NFS client and a 4.0/i386 NFS
server but this is reproductible with every box I tried so far (mostly
current/i386).
$ du -ch file
44.4M file
44.4M total
# NFSv3
$ sudo mount_nfs mcp:/export /mnt/nfs
$ mount -v |grep /mnt/nfs
mcp:/export on /mnt/nfs type nfs (rw, ctime=Tue Feb 6 16:37:33 2007,
v3, udp, hard, wsize=8192, rsize=8192, rdirsize=8192, timeo=100,
retrans=10, maxgrouplist=16, readahead=1, acregmin=5, acregmax=60,
acdirmin=5, acdirmax=60)
$ time cp file /mnt/nfs/
0m7.90s real 0m0.01s user 0m0.35s system
$ sudo umount /mnt/nfs
# NFSv2
$ sudo mount_nfs -2 mcp:/export /mnt/nfs
$ mount -v |grep /mnt/nfs
mcp:/export on /mnt/nfs type nfs (rw, ctime=Tue Feb 6 16:41:38 2007,
v2, udp, hard, wsize=8192, rsize=8192, rdirsize=8192, timeo=100,
retrans=10, maxgrouplist=16, readahead=1, acregmin=5, acregmax=60,
acdirmin=5, acdirmax=60)
$ time cp file /mnt/nfs/
0m57.49s real 0m0.00s user 0m0.42s system
While those values are not meant to be very accurate, they do show a
problem.
Am I the only one in this situation? Any hidden tricks to improve write
transfer speed?
Cheers!
--
Antoine
