G1 parallel old gen collection

charlie hunt charlie.hunt at oracle.com
Tue Oct 4 20:34:59 UTC 2016


> On Oct 4, 2016, at 3:27 PM, Christian Thalinger <cthalinger at twitter.com> wrote:
> 
>> 
>> On Oct 4, 2016, at 10:22 AM, charlie hunt <charlie.hunt at oracle.com> wrote:
>> 
>> Comments below.
>> 
>>> On Oct 4, 2016, at 3:11 PM, Christian Thalinger <cthalinger at twitter.com> wrote:
>>> 
>>>> On Oct 4, 2016, at 9:45 AM, charlie hunt <charlie.hunt at oracle.com> wrote:
>>>> 
>>>>> Is that still the case in JDK 9?
>>>> 
>>>> Yes
>>> 
>>> What’s the answer to my second question?
>> 
>> It’s in the backlog (not actively being worked on).
> 
> Ok.
> 
>> 
>>> 
>>>> 
>>>> We need to figure out why you’re getting full GCs, and what to do to avoid them.
>>>> 
>>>> If you can share the GC logs from G1, and the command line options in use, we can probably help improve the situation for you. We will likely suggest enabling +PrintAdaptiveSizePolicy so we can see why G1 is making certain adaptive decisions.
>>> 
>>> I can certainly turn that on.  On sharing logs, I have to check if that’s possible.  Also, keep in mind this is JDK 8u.
>> 
>> JDK 8u is fine. G1 is tunable in JDK 8u. ;-)
>> 
>> Btw, knowing you’re an adventurous guy :-] … If you can get your hands on a copy of the “Java Performance Companion”, there’s several chapters in it that cover in detail G1 including how to go about tuning it  (assuming you wanna know more details, or interested in going down the DIY path).
> 
> Well, it’s not that we don’t have G1 expertise in-house ;-)

Yep!  Tony was a reviewer of that content. :-)

> 
> Sure, avoiding full GCs should be the goal but if that isn’t possible (for some reason) it might make sense to make them parallel.  We wanted to double-check with Oracle so we are not duplicating work here.
> 
> Still, I’d like to get some logs to you anyway, if possible.

Sounds great!

charlie

> 
>> 
>> hths,
>> 
>> charlie
>> 
>>> 
>>>> 
>>>>> On Oct 4, 2016, at 2:37 PM, Christian Thalinger <cthalinger at twitter.com> wrote:
>>>>> 
>>>>> First, I apologize if I get all the terminology wrong.  You know me… this is all a blackbox to me.
>>>>> 
>>>>> I’m currently trying to run some Twitter services with G1 and the results are encouraging but not good enough.  One issue is that full GCs take much longer than with ParallelOld (3-4x longer).
>>>>> 
>>>>> Tony said that G1’s old collection is not parallel.
>>>>> 
>>>>> Is that still the case in JDK 9?  If yes, is someone at Oracle working on parallelizing it?

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.java.net/pipermail/hotspot-gc-dev/attachments/20161004/b16cf8d6/attachment.htm>


More information about the hotspot-gc-dev mailing list