Performance problems with NFS may be attributable to various causes:
Insufficient disk performance on the server can limit the throughput rate at which reads and writes can be performed.
- Excessive CPU load on client or server limits their capability to service network requests.
A congested network can limit the transfer rate or increase the retry rate for requests or data transfers.
Improving read and write access
If the data transfer for an NFS read or write access was not successfully completed, then the entire data block is transferred once again. If the retry rate is too high, the size of the NFS read or write blocks must be reduced. This can be done by setting the option rsize and wsize in the file /etc/vfstab or, during mounting, by using the mount command.
Try choosing the value 2048 bytes or 1024 bytes for the entry in the file /etc/vfstab:
bobserver:/home/bob - /home/bob nfs - - rw,rsize=2048,wsize=2048
Since a reduction in size of the read and write blocks restricts the maximum possible performance, it may also be advantageous to reduce the time limit for the request retries. Set the option timeo for the mount command or in the file /etc/vfstab to 8 or 6 tenths of a second (the default is 11 tenths of a second).
Improving network performance
If a mounted NFS file system extends over a wide network encompassing many gateways or great distances between clients and server, it may be necessary to increase the option timeo for the mount command. Slow network hardware can cause long delays in the transmission path. The time limit may need to be increased in this case. For example, set timeo to 50 or 100 tenths of a second.
Use nfsstat -rc to check the number of NFS/RPC transfer retries and errors affecting the client.