ParNew and CMS-Initial Corelation

I noticed that when there is no recent ParNew, and then, if the initial CMS-Initial-Mark phase begins, the CMS-Initial-Mark phase takes longer to mark objects in the old generation.

The good part is most of the time when I found ParNew (maybe by accident or the JVM does it inside) happens just before the CMS-Initial-Mark phase, and then to indicate the same number of objects in the old generation, CMS takes less time.

I would like to know the reasons for this observation.

Note. Given that the CMS-Initial-Mark phase is a stop world, you should try to reduce its duration.

+3
source share
1 answer

CMS . CMS (2 ). , , .

http://blog.ragozin.info/2011/06/understanding-gc-pauses-in-jvm-hotspots_02.html.

HotSpot -XX: CMSWaitDuration = < delay in ms > , CMS .

+5

All Articles