Posted by larst
on March 27, 2009 at 12:59 AM PDT
Testdriving the new Garbarge-First (G1) on an Enterprise T5220 looks promissing in the beginning, but once a full GC is needed everything halts for ages.
Examination of the CPU utilization shows only one out of 64 CPU threads beeing utilized during the full GC operation. So no wonder this operation takes time.
At the moment I am using these settings:
-server -Xmx3072m -XX:MaxPermSize=256m -XX:+UnlockExperimentalVMOptions -XX:+UseG1GC
I have tried experimenting with the -XX:ParallelGCThreads setting, but G1 does not seem to honor this setting.
Given the rather sparse information on using the G1 collector I hope somebody here is able to point me in the right direction on how to utilize the rest of the CPU during this stop-the-world phase...