The tests conducted with the number of users increased from 1 to 100 users, stepping up by 20 users at a time, with no think time between requests from the load driver. This allowed quick turnaround time per test run, and created results that were easier to present. The same loads were run using 1 second think times, but it scales to so many users that it makes it difficult to present the data. The results were very similar anyway, from a throughput improvement perspective. Before each test run the order transactions in the workload were deleted and the InnoDB buffer pool was checked for dirty pages before starting the next run. Each test was run five times to ensure reproducibility of the results. A maximum of 100 users was used because at was at that point which throughput plateaued.