G1 feedback: frantic GC cycles

Simone Bordet sbordet at intalio.com
Wed May 4 15:28:38 UTC 2011


On Wed, May 4, 2011 at 16:20, Tony Printezis <tony.printezis at oracle.com> wrote:
> Igor,
> FWIW, 518M->517M doesn't necessarily mean that we only filled on region. It
> might mean that we filled, say, 5 but most objects survived. Having said
> that, given how the previous GCs behaved, I think you're right that we
> probably only collected one region. I'm quite perplexed with why we went
> from collecting 256M of young gen to 1M (!!!):
>  [ 760M->506M(1024M)]
>  [ 507M->506M(1024M)]
>  [ 507M->506M(1024M)]
>  [ 507M->506M(1024M)]
> Simone, you use 1g heap, right? Can you maybe add / set the following
> params:
> -Xms1g -Xmx1g -Xmn256m
> -XX:-UseAdaptiveSizePolicy

Disabling heap resizing, ok.

Would it be possible that resizing of the young generation is way too
aggressive in G1 ?

> -XX:InitiatingHeapOccupancyPercent=60

Tony, would you mind to explain this parameter ?
I ask to increase my knowledge about G1 :)

Thanks !

Finally, no matter how good the architecture and design are,
to deliver bug-free software with optimal performance and reliability,
the implementation technique must be flawless.   Victoria Livschitz

More information about the hotspot-gc-dev mailing list