Testing & Tuning Hyperconverged OpenStack

OpenStack clusters are most often built with servers for Nova compute VMs and servers for storage, with Ceph storage requring 3 or more nodes. It can be more cost-effective to “hyperconverge” Nova and Ceph on to the same servers, and rising processor core counts and RAM density have made this feasible. But it is important to understand the resource demand patterns of each and protect against corner cases where one starves the other. We will present an empirical approach to: Generating realistic

source by Open Infrastructure Foundation

openstack