https://bz.apache.org/bugzilla/show_bug.cgi?id=58786
--- Comment #17 from Felix Schumacher <[email protected]> --- Those results seem to be more correct. If you look at profiling data (I obtained them with java 8 and jmc), you will see, that the actual time spent in the db pool is minimal compared to the time spent in the db or in other parts of jmeter. I used JVM_ARGS=-"XX:+UnlockCommercialFeatures -XX:+FlightRecorder -XX:FlightRecorderOptions=defaultrecording=true,dumponexit=true,dumponexitpath=/tmp/flightrecording.jfr" ./bin/jmeter -n -t ~/BUG_58786.jmx to record the profiling data and used jmc to look at them. Looking at your numbers I don't see a reason to update the implementation for its speed. If we want to swap implementation because of excaliburs age (and given it is in attic), then I would favour dbcp2, even if it is a bit slower, because it has more developers than tomcat-jdbc and hikari. -- You are receiving this mail because: You are the assignee for the bug.
