For request-based workloads, report request throughput separately
steveblackburn opened this issue · comments
Some request based workloads have a preparation phase before commencing request processing. The preparation phase is part of the overall benchmark, but is unrelated to request processing. When studying request processing, it would be useful to be able to report the time for the period in which requests are processed (separate from any preparatory work). Hooks to do this already exist, but at the moment the request based workloads only print out request latency metrics.
Possibly a better solution is to optionally redefine the benchmark start and end as being the start and end of request processing. A command line flag would lead to only that period of request-based workloads being considered as part of the workload. Other workloads would be unaffected by the flag.
Added a line to the output that reports throughput:
===== DaCapo processed 524288 requests in 20885 msec, 25103 requests per second =====