We have two networks in Ceph. Public network on 1Gbit/s and Cluster network on 10Gbit/s. But I don't see that clients use whole 1Gbit bandwith interface in OSD node. We use Ceph storage for OpenStack. And instances in OpenStack have very low write and read data. For storage We have 6 servers HP DL380 G9 with 14x900GB, 3 servers HP DL380 G9 with 14x900Gb and 4x1.8Tb and 2 servers Huawei RH2288H v3 14x900Gb for storage, all nodes have 4 SSD for journals and 3 monitors run on a together with OSD nodes. But in the future We will move them to separate hosts. We use Ceph version Hammer, but one node in version Jewel. At the moment I do update storage to Luminous. Please help, I can't find bottleneck in our storage cluster.