I've got two Ubuntu machines, an NFS server running Ubuntu 10.10 2.6.32-25 and a client running 10.04 2.6.32-24. The server is exporting a 12tb RAID-6 locally attached array (it's a Dell PE R515). It's got nfs-common and nfs-kernel-server 1.2.0-4 installed, along with portmap-6.0.0-1.

/etc/exports is thus:
/data/nfs xxx.xxx.xxx.xxx/26(rw,sync,no_root_squash,no_subtree_check)

fstab on the client is thus:
nfsserver:/data/nfs /mnt/nfs nfs rw,hard,intr 0 0

The mount works fine, but when copying data to/from the NFS share from the client, my shell just grinds to a halt. If I open a new window on the client and try to navigate directories on the nfs server and do something simple like an ls, it takes minutes to execute. I've got other nfs installs that work just fine. I tried switching to nfs4, with the same behavior. I've ruled out network problems (duplexing, etc).. ftp transfers are speedy.. even if the throughput was slow, my shell shouldn't hang like this so I'm tempted to blame software. If I copy more than 1 file concurrently, forget it- I might as well go get coffee.

Any ideas?

Other relevant info:

root:~# cat /etc/services |grep 111
sunrpc 111/tcp portmapper # RPC 4.0 portmapper
sunrpc 111/udp portmapper
kx 2111/tcp # X over Kerberos
root:~# cat /etc/services |grep 2049
nfs 2049/tcp # Network File System
nfs 2049/udp # Network File System
root:~# rpcinfo -p
program vers proto port
100000 2 tcp 111 portmapper
100000 2 udp 111 portmapper
100024 1 udp 36732 status
100024 1 tcp 55080 status
100021 1 udp 43335 nlockmgr
100021 3 udp 43335 nlockmgr
100021 4 udp 43335 nlockmgr
100021 1 tcp 44712 nlockmgr
100021 3 tcp 44712 nlockmgr
100021 4 tcp 44712 nlockmgr
100003 2 udp 2049 nfs
100003 3 udp 2049 nfs
100003 4 udp 2049 nfs
100003 2 tcp 2049 nfs
100003 3 tcp 2049 nfs
100003 4 tcp 2049 nfs
100005 1 udp 46312 mountd
100005 1 tcp 51940 mountd
100005 2 udp 46312 mountd
100005 2 tcp 51940 mountd
100005 3 udp 46312 mountd
100005 3 tcp 51940 mountd

You could try adding rsize and wsize options in your NFS exports. The defaults might not be optimal for your data. You would need to experiment with the sizes to see if you notice a performance difference. Additional, you could take out the sync option, which would increase performance at the cost of data validation (which increases read/write overhead). Of course that is not advisable, if you are concerned about data corruption (not common, but still possible).