I have been using YCSB for benchmarking, but since YCSB is overly complicated I have now looked at Pillowfight (cbc-pillowfight) to compare how, with less CPU cycles for the loads, it’s possible to run workloads similar to YCSB for a better comparison between different workloads, runtimes, number of threads etc.
Doing this, out of the box, pillowfight seem to have some limitations mostly related to the limited output with regards to analyzing the results from the runs. I would like to ask if Couchbase or any users have any pre-maid scripts to get reports from the output logs to read the IOPS(average - maybe even min and max per time cycle), the read vs. update latencies, average latencies, 95/99% latencies at least. Some of these details can be calculated from the output files, but since the output log becomes fairly messy when using more than one thread I’d like to ask if someone has already gone through the hazzle to create scripts for this type of reporting - and would be willing to share?
Alternatively - I’m also open to a pointer for a Benchmarking tool which reports these details in a less CPU consuming way than YCSB?