Effects of NFS data caching on write throughput

If you are trying to perform sequential write operations on files using NFS Version 3 or NFS Version 4 that are larger than client memory, you can improve performance by using commit-behind.

Writing entire files that are larger than the amount of memory in the client causes heavy page replacement activity on the client. This might result in a commit operation being performed over-the-wire for every page of data written. Commit-behind enables a more aggressive logic for committing client pages to stable storage on the server and, more importantly, returning those pages to the free list.

You can enable commit-behind when mounting the file system by specifying the combehind option with the mount command. You also need to set an appropriate value for the numclust variable, with the mount command. This variable specifies the number of 16 KB clusters processed by the sequential write-behind algorithm of the Virtual Memory Manager (VMM). When the I/O pattern is sequential, use a large value for the numclust option in order to keep more pages in RAM before scheduling them for I/O. Increase the value for the numclust option if striped logical volumes or disk arrays are being used.