0

Is it advisable to switch from CMS to G1 with large object requests(approx. 40-50MB). Available resource details: Heap Size : Between 6GB-16GB. Heap region size: 8MB-32MB. I applied G1 GC with following configurations: Heap Size: 12GB, Region size: 8MB but due to large no. of Humongous object requests triggers Full GC was triggered(After checking the logs I found out that humongous objects were not getting reclaimed because the initial-mark initiated during humongous allocation never got finished which results in old generation getting completely filled). Should I stick to G1 or roll back to CMS?

toshi1894
  • 1
  • 1
    Why don't you benchmark both GC's with typical stress loads from your application? – Eric J. Sep 04 '15 at 04:01
  • Hey @EricJ. actually a lot of applications are running on my host. I tried CMS there is a lot of fragmentation. To avoid this I switched to G1. I just want to know whether G1 can handle such large request if they come very frequently ? – toshi1894 Sep 04 '15 at 06:14
  • see [this answer](http://stackoverflow.com/a/28065423/1362755) for references regarding humongous objects and G1 – the8472 Sep 04 '15 at 09:06

0 Answers0