Incremental load testing is a good way to measure system performance because workload is added in steps, with metrics collected for each step. Using an incremental load-testing approach can also enable the development of a performance model based on the trends produced by the step metrics. Incremental load testing works by running a load test for a specific workload for a time that is needed for the system to reach a steady state of workload execution. When the first step-load test is complete, the load is increased and held again until the system stabilizes. The trends that result from all or some of the steps can then be used to model performance of additional workload on the same system.
The following figure illustrates the incremental load-test methodology.
Figure 1. Example of an incremental load-test methodology
We developed a test profile that worked across both the legacy and modern hyper-converged platforms. A workload configuration that exerted a workload that was too large might work well on the modern hyper-converged platform but not on the legacy system. For this reason, we simulated a small database workload. We used Quest Benchmark Factory to simulate users and transactions, and we used the following TPC-E-like configuration:
For configuration details about the S2D Ready Nodes, legacy system, and SQL Server databases that we used in our testing, see Appendix A: Detailed configurations.