It has long been said that data expands to fill available space, a phenomenon quite familiar to Ceph admins. As your constellation of Ceph clients continues to change and grow, it becomes increasingly vital to plan for sustained performance. Many factors at the operating system and Ceph architectural levels become increasingly important as Ceph clusters grow in capacity and workload.
Fortunately, Ceph is up to the challenge, and we will explore a number of ways to measure and optimize your deployments in this final chapter.
As we've explored in earlier chapters, many traditional and commercial storage systems exhibit performance that is limited by a single SAS, Fiber Channel, or network interface. As they are scaled for capacity, throughput may remain constant, or at best grow at a lower rate than capacity. Among Ceph's strengths is the ability to avoid data-plane choke points so that we are able to scale capacity and throughput together as we add additional storage...