The metric Garbage Collection Time Spent (ms/min) is very indicative of the health of the heap configuration. I'd like some clarification fot understanding its implication.
Both Garbage Collection Time and CPU time can be real-time and cpu-time. For a 10-core server, the total cpu-time is 1000% and the GC time can be 400% briefly if 4 cores are used for parallel GC.
For the numerator, GC Time, is it real-time or cpu-time, both in the case of concurrent non-pause phase and parallel and non-parallel pause phase?
For the denominator, the total time, is the one minute real-time clock time, or real-time multiplied by the core count?
That affects the interpretation significantly. For example, 20000 ms/min looks very bad, 33%. But if most of the GC time is on concurrent mark, 20 secs one-core cpu-time is not bad.