This document describes large-scale I/O characterization testing performed by the Red Hat Performance & Scale Engineering group. It uses Red Hat OpenStack Platform (RHOSP) 10 with Glance, Nova, and Cinder on 1.8 PB of Red Hat Ceph Storage (RHCS) 2.0 storage. This RHOSP configuration used a pre-existing external RHCS cluster of 29 servers and 20 RHOSP compute nodes, possibly the largest Red Hat internal RHCS deployment to date. Up to 512 VMs were exercised measuring application latency percentiles as a function of time, while operational events including OSD failure, OSD node failure, and Monitor failure were simulated.
The purpose of this testing was to discover any problems with the RHOSP-RHCS configuration at scale as well as to document and understand the user experience for a configuration of this type and size with respect to performance, and specifically to understand issues customers may experience with such a configuration in production. This effort resulted in the filing of one RHCS and five RHOSP problem reports.
- Article Type
- Performance Brief