Optimizing performance

With the volume created and mounted, we can tweak some parameters to get the best performance. Mainly, performance tuning can be done on the filesystem level (in this case, ZFS), and on the GlusterFS volume level.

GlusterFS tuning

Here, the main variable is performance.cache-size. This setting specifies the amount of RAM to be allocated as a read cache for the GlusterFS volume. By default, it is set to 32 MB, which is fairly low. Given that the selected VM has enough RAM, this can be bumped to 4 GB using the following command:

gluster volume set gvol1 performance.cache-size 4GB

Another essential parameter once the cluster starts growing is performance.io-thread-count. This controls how many I/O threads...