Storage Benchmarking for Workload Aware Storage Platform
Wubin Li, Tremblay Benoit · 2016
Storage backend characterization is vital to the workload aware storage platform which aims at providing optimal mapping between storage workloads and backends. Storage benchmarking as one major activity in the path of backend characterization, is very challenging due to the diversity and complexity of the storage environments. In this paper, we present our storage benchmarking design and implementation for the workload aware storage platform. A set of workload classes each of which consists of multiple IO workloads are defined to represent application scenarios. By varying the intensity parameter that indicates the scale of workloads in workload classes, our approach examines the performance of storage backends in a black-box manner, without requiring internal knowledge of the storage backends. The applicability of the proposed solution is evaluated using a video streaming case study on an OpenStack cluster with Ceph as the storage backend.