Slow NFS performance may indicate problems with nfsd daemons on the server or biod daemons on the client. The number of default nfsd and biod daemons (four of each are started) can be changed by editing these numbers where these daemons are invoked in the /etc/nfs file. If there is heavier traffic across the network than normal or if the network traffic is very spread out, the number of default nfsd daemons can be increased to speed up the system.
If access to remote files seems unusually slow,
the server should be checked by entering (on the server):
ps -ef | more
If the server is functioning and other users are
getting good response, use the ps -ef
command on the client to see whether block I/O
(biod) daemons are running.
If the daemons are not running or are hung,
they should be killed. First, find the process
IDs by entering:
ps -ef | grep biod
Then, kill them with:
kill -9 pid1 pid2 pid3 pid4
Restart the daemons with:
To determine whether the daemons are hung, use ps as above, then copy a large file. Another ps shows whether the biods are accumulating CPU time; if not, they are probably hung.
If biod appears to be functioning correctly, check the network connection. Use the nfsstat -c and nfsstat -s commands to discover whether a client is doing a lot of retransmitting. A retransmission rate of 5% is considered high. Excessive retransmission usually indicates a bad network adapter, a bad network tap, a mismatch between adapter and tap, or a mismatch between the client machine's network adapter and the server's adapter. Insufficient STREAMS buffers may also cause excessive retransmissions. For more information, see the Performance Guide.