Having ample network bandwidth is crucial for Ceph or any other storage. The more bandwidth that is dedicated, the more VM's performance and latency will benefit. Note here that when a dedicated journal such as SSD is used, the requirement for network bandwidth will increase significantly, because more data will traverse the Ceph cluster for replication and distribution. For a Ceph cluster where SSD is used as a dedicated journal, a gigabit network should not be used for the Ceph cluster network. At the very least, 10 GB would be a good network. We can also use InfiniBand as an alternative network solution on a lower budget. If neither is possible, then multiple bonded gigabits would also work. On a single gigabit, the network will become a bottleneck, causing cluster-wide performance degradation.

Also, Ceph cluster sync should be on its own dedicated network, with the Ceph public facing network on another. Ceph uses the cluster network to commit all syncs between OSDs. This prevents unnecessary load on the public facing network.