How does JVM decide the tenuringThreshold?

Thu Dec 17 19:55:49 UTC 2015

Hi Jun,

Ok, I see. As Jon mentioned, Parallel GC's survivor policy is quite
different than other collectors. Rather, it uses two sets of heuristics
mechanism:

(1) Balance of young gc and full gc percentage to determine whether to
decrease/increase tenuring_threshold.
(2) Check if survivor overflows during young gc or if desired target size
is too large under current tenuring_threshold to decrease the threshold.

Other collectors scans ageTable to determine optimal threshold to honor
TargetSurvivorRatio ("Desired" survivor object size / survivor space).

Turning on -XX:+PrintAdaptiveSizePolicy would output more info about
Parallel GC tenuring threshold adjustment.

Thanks.
Tao Mao

On Thu, Dec 17, 2015 at 7:08 AM, Jun Zhuang <jun.zhuang at hobsons.com> wrote:

> Hi,
>
>
>
> The JVM options for the 1st log is: -Xms2g -Xmx2g -XX:MaxPermSize=256m
> -XX:NewSize=128m -XX:MaxNewSize=128m -XX:SurvivorRatio=126. But since
> adaptive sizing policy is in play, most of the setting will be ignored
> anyway.
>
> The JVM options for the 2nd log is: -Xms2g -Xmx2g -XX:MaxPermSize=256m
> -XX:NewSize=512m -XX:MaxNewSize=512m -XX:SurvivorRatio=510
> -XX:-UsePSAdaptiveSurvivorSizePolicy
>
> Thanks,
>
> Jun
>
>
>
> *From:* Tao Mao [mailto:yiyeguhu at gmail.com]
> *Sent:* Wednesday, December 16, 2015 5:11 PM
> *To:* Jun Zhuang <jun.zhuang at hobsons.com>
> *Subject:* Re: How does JVM decide the tenuringThreshold?
>
>
>
> Could you provide your JVM options for the 1st and 2nd logs? Thanks. -Tao
>
>
>
> On Wed, Dec 16, 2015 at 1:52 PM, Jun Zhuang <jun.zhuang at hobsons.com>
> wrote:
>
> Hi,
>
> I am trying to gain some knowledge regarding how GC works by running load
> tests against a java web service, hosted on JBOSS, and adjusting various
> java startup parameters. One interesting thing I noticed was the tenuring
> threshold, I thought it would change with size of the survivor space, but I
> could never make it happen, can someone provide some clue? Following is
> what I did:
>
> *JAVA startup parameters:*
>
> -Xms2g -Xmx2g -XX:MaxPermSize=256m -XX:NewSize=128m -XX:MaxNewSize=128m
> -XX:SurvivorRatio=126
>
> I then changed the SurvivorRatio to 30, 6, and 1, but the “new threshold”
> in the GC log always stayed at 1 after the first few young GCs, adding
> -XX:TargetSurvivorRatio=90 did not seem to make any difference:
>
>  4.504: [GC
>
> Desired survivor size 16777216 bytes, new threshold 7 (max 15)
>
>  [PSYoungGen: 98304K->7553K(114688K)] 98304K->7553K(2080768K), 0.0199500
> secs] [Times: user=0.04 sys=0.01, real=0.02 secs]
>
> 12.784: [GC
>
> Desired survivor size 16777216 bytes, new threshold 7 (max 15)
>
>  [PSYoungGen: 105857K->8979K(114688K)] 105857K->8979K(2080768K), 0.0177280
> secs] [Times: user=0.04 sys=0.01, real=0.01 secs]
>
> 18.935: [GC
>
> Desired survivor size 16777216 bytes, new threshold 7 (max 15)
>
>  [PSYoungGen: 107283K->13322K(114688K)] 107283K->13322K(2080768K),
> 0.0282660 secs] [Times: user=0.06 sys=0.00, real=0.02 secs]
>
> 22.944: [GC
>
> Desired survivor size 16777216 bytes, new threshold 7 (max 15)
>
>  [PSYoungGen: 111626K->11304K(114688K)] 111626K->11377K(2080768K),
> 0.0255770 secs] [Times: user=0.08 sys=0.00, real=0.03 secs]
>
> 24.089: [GC
>
> Desired survivor size 16777216 bytes, new threshold 6 (max 15)
>
>  [PSYoungGen: 109608K->16377K(114688K)] 109681K->17678K(2080768K),
> 0.0373650 secs] [Times: user=0.10 sys=0.01, real=0.04 secs]
>
> 24.801: [GC
>
> Desired survivor size 16777216 bytes, new threshold 5 (max 15)
>
>  [PSYoungGen: 114681K->16358K(114688K)] 115982K->21621K(2080768K),
> 0.0627800 secs] [Times: user=0.21 sys=0.01, real=0.06 secs]
>
> 25.500: [GC
>
> Desired survivor size 16777216 bytes, new threshold 4 (max 15)
>
>  [PSYoungGen: 114662K->16359K(114688K)] 119925K->30181K(2080768K),
> 0.0577540 secs] [Times: user=0.17 sys=0.01, real=0.06 secs]
>
> 26.280: [GC
>
> Desired survivor size 16777216 bytes, new threshold 3 (max 15)
>
>  [PSYoungGen: 114663K->16354K(114688K)] 128485K->37506K(2080768K),
> 0.0539360 secs] [Times: user=0.15 sys=0.01, real=0.05 secs]
>
> 27.330: [GC
>
> Desired survivor size 16777216 bytes, new threshold 2 (max 15)
>
>  [PSYoungGen: 114658K->14928K(114688K)] 135810K->42581K(2080768K),
> 0.0588730 secs] [Times: user=0.20 sys=0.01, real=0.06 secs]
>
> 28.108: [GC
>
> Desired survivor size 16777216 bytes, new threshold 1 (max 15)
>
>  [PSYoungGen: 113232K->16363K(114688K)] 140972K->59269K(2080768K),
> 0.0586460 secs] [Times: user=0.15 sys=0.01, real=0.06 secs]
>
> . . .
>
> 1538.726: [GC
>
> Desired survivor size 3932160 bytes, new threshold 1 (max 15)
>
>  [PSYoungGen: 129280K->1792K(125184K)] 433305K->308417K(2091264K),
> 0.2403020 secs] [Times: user=0.83 sys=0.01, real=0.24 secs]
>
> 1543.537: [GC
>
> Desired survivor size 4063232 bytes, new threshold 1 (max 15)
>
>  [PSYoungGen: 125184K->2592K(126016K)] 431809K->310233K(2092096K),
> 0.2318620 secs] [Times: user=0.83 sys=0.01, real=0.23 secs]
>
> 1548.268: [GC
>
> Desired survivor size 3932160 bytes, new threshold 1 (max 15)
>
>  [PSYoungGen: 125984K->1344K(127232K)] 433625K->310657K(2093312K),
> 0.2294920 secs] [Times: user=0.83 sys=0.01, real=0.23 secs]
>
>
>
>
>
> If the threshold is really 1, then nothing should be promoted to the old
> gen, but that definitely was not the case.
>
>
>
>
> -------------------------------------------------------------------------------------------------------
>
>
>
> I also experimented with -XX:-UsePSAdaptiveSurvivorSizePolicy, the “new
> threshold” always stays at 7 despite the SurvivorRatio value.
>
>
>
> 1643.408: [GC
>
> Desired survivor size 4194304 bytes, new threshold 7 (max 15)
>
>  [PSYoungGen: 519729K->4032K(520192K)] 935797K->420443K(2093056K),
> 0.2605620 secs] [Times: user=0.94 sys=0.01, real=0.26 secs]
>
> 1662.092: [GC
>
> Desired survivor size 4194304 bytes, new threshold 7 (max 15)
>
>  [PSYoungGen: 520128K->4096K(520192K)] 936539K->421235K(2093056K),
> 0.2588890 secs] [Times: user=0.95 sys=0.01, real=0.26 secs]
>
> 1680.882: [GC
>
> Desired survivor size 4194304 bytes, new threshold 7 (max 15)
>
>  [PSYoungGen: 520192K->4096K(520192K)] 937331K->422347K(2093056K),
> 0.2661180 secs] [Times: user=0.97 sys=0.00, real=0.27 secs]
>
> 1698.975: [GC
>
> Desired survivor size 4194304 bytes, new threshold 7 (max 15)
>
>  [PSYoungGen: 520192K->4096K(520192K)] 938443K->423246K(2093056K),
> 0.2626090 secs] [Times: user=0.95 sys=0.01, real=0.26 secs]
>
>
>
>
>
> Appreciate any input,
>
> Jun
>
>
>
> *Jun Zhuang*
>
> *Sr. Performance QA Engineer | Hobsons*
>
>
>
>
> _______________________________________________
> hotspot-gc-use mailing list
> hotspot-gc-use at openjdk.java.net
> http://mail.openjdk.java.net/mailman/listinfo/hotspot-gc-use
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/hotspot-gc-use/attachments/20151217/c416a539/attachment-0001.html>