Any software requires CPU cycles, and GlusterFS is no exception. CPU requirements are moderately low, and depend on the type of volume used, for example, a replicated volume requires far less CPU than a dispersed volume.
CPU requirements are also affected by the type of filesystem that the bricks use and what features they have. Going back to the ZFS example, if compression is enabled this adds increased load to the CPU, and not having enough CPU resources decreases performance considerably.
For a simple storage server and no advanced features at the brick level, anything with four CPUs or more is sufficient. When enabling filesystem features, such as compression, eight or more CPUs are required for optimal performance. Additionally, more CPU allows for more concurrent I/O to be done to the cluster. This is of utmost importance when designing a storage cluster for high-performance compute (HPC) applications, where thousands of users are performing I/O operations at the same time.
- For highly concurrent workload, go for higher CPU count, above eight CPUs, depending on the concurrency level
- For low-performance requirements and a cost-efficient solution, select a lower number of CPUs, for example, four CPUs