Characterizing the performance and resilience of HCI clusters with the TPCx-HCI benchmark

Abstract

We use the newly-released TPCx-HCI benchmark to characterize the performance and resilience properties of Hyper-Converged Infrastructure clusters. We demonstrate that good performance on an HCI cluster requires delivering all properties of high IOPS, low latencies, low CPU overhead, and uniform access to data from all Nodes. We show that unless the cluster can quickly and efficiently rebalance the VMs after a change in the workload, performance will be severely impacted. We use the data accessibility test of TPCx-HCI to show how performance is impacted by rebuilding traffic after a Node goes down, and how long it takes for the rebuilding to finish.