Use the following general guidelines to configure shared file systems:
Make sure the network has enough bandwidth.
Make sure the underlying storage has enough I/O bandwidth.
Configure the shared file system daemons, particularly the client, to have enough threads to access files quickly. For example, IBM recommends that you estimate the number of files that require simultaneous access and provide at least two biod threads for each file.
When you run concurrent sessions on a grid that use flat file sources or targets, provide enough threads so each partition can access the source or target files that they need simultaneously.
Configure mount points of the shared file system based on access requirements. When running sequential sessions on a grid that use flat file sources or targets, avoid any configuration that might degrade the effectiveness of the default read-ahead or write-behind process. File systems optimize sequential file access with read-ahead and write-behind.
If necessary, tune the shared file system read-ahead and write-behind settings.
Review the cache settings of the shared file systems for both the client and server. Increasing the default settings may improve performance.
Configure the release-behind settings of the file system to free memory pages after data is accessed. Otherwise, system performance might degrade when reading or writing large files.
Because of the difference in access patterns, you might use different mount points for sources and targets, and persistent caches.
For more information, see the shared file system documentation.