Unexpected Java 6 vs 8 GC performance difference

Mon Jun 6 23:08:55 UTC 2016

Jon,

I don't have a very good benchmark, unfortunately, but I've attached the program
I used to convince myself I wasn't crazy to suspect the interned strings. It's a simple
program that just interns 5 million strings, then spams a bunch of object creations
so the GC run a lot.

Note that while running it with Java 8 will actually result in the program running quicker,
if you look at the GC logs produced, you'll note that minor GCs are taking around 7x longer
in Java 8 on average.

You'll probably want to tweak settings yourself, but if not, or as a starting point, there's a
run.sh script attached that can be used to run it like ./run.sh java_version builtin|map
(e.g. "./run.sh 6 builtin" or "./run.sh 8 map").

Please let me know if I can do anything else to help. I'd love to get to the bottom of why
interned strings take so long to collect, especially when none of them are actually being
freed, though I suspect the reason is simply that no one ever expected anyone to intern
millions of strings and that it's a very rare case not worth optimizing.

Thanks,
Corbin

From: hotspot-gc-use [mailto:hotspot-gc-use-bounces at openjdk.java.net] On Behalf Of Jon Masamitsu
Sent: Friday, June 03, 2016 11:37 AM
To: hotspot-gc-use at openjdk.java.net
Subject: Re: Unexpected Java 6 vs 8 GC performance difference

Corbin,

Do you have a benchmark that will exhibit this problem?
If not, does you application just create lots of interned strings that are long
lived?

Thanks.

Jon
On 5/31/2016 3:36 PM, Hughes,Corbin wrote:
After coming across -XX:+PrintGCTaskTimeStamps, I noticed that the listed tasks don't add up to the listed PSYoungGen times, so I ended up adding more "printf" style timing outputs to the hotspot source, and ended up tracking it down. Turns out it's the interned strings. The way they're GC'd isn't particularly efficient for large numbers of strings: the call to StringTable::unlink_or_oops_do(&_is_alive_closure, &root_closure); was taking ~185 ms of the ~200 ms. I wish there were a flag to disable GC of interned strings, but I think we'll be able to get away with using something like a Guava Interner instead of the built in StringTable, so mystery solved.

From: hotspot-gc-use [mailto:hotspot-gc-use-bounces at openjdk.java.net] On Behalf Of Hughes,Corbin
Sent: Friday, May 27, 2016 9:24 PM
To: hotspot-gc-use at openjdk.java.net<mailto:hotspot-gc-use at openjdk.java.net>
Subject: Re: Unexpected Java 6 vs 8 GC performance difference

Jon,

Thanks for the interest. Hopefully this log will be more helpful. This is Java 7 with the -XX:-UseParallelOldGC flag. Perhaps notably, the PSYoungGen collections are even slower in Java 7 (regardless of whether or not the serial old collector is used).

Thanks,

Corbin

{Heap before GC invocations=456 (full 0):
PSYoungGen      total 2091008K, used 2087456K [0x0000000780000000, 0x0000000800000000, 0x0000000800000000)
  eden space 2084864K, 100% used [0x0000000780000000,0x00000007ff400000,0x00000007ff400000)
  from space 6144K, 42% used [0x00000007ff400000,0x00000007ff688000,0x00000007ffa00000)
  to   space 6144K, 0% used [0x00000007ffa00000,0x00000007ffa00000,0x0000000800000000)
PSOldGen        total 3145728K, used 2267941K [0x00000006c0000000, 0x0000000780000000, 0x0000000780000000)
  object space 3145728K, 72% used [0x00000006c0000000,0x000000074a6c9508,0x0000000780000000)
PSPermGen       total 819200K, used 57100K [0x0000000680000000, 0x00000006b2000000, 0x00000006c0000000)
  object space 819200K, 6% used [0x0000000680000000,0x00000006837c3130,0x00000006b2000000)
1461.284: [GC
Desired survivor size 6291456 bytes, new threshold 1 (max 15)
[PSYoungGen: 2087456K->1482K(2091008K)] 4355397K->2269463K(5236736K), 0.2986350 secs] [Times: user=0.34 sys=0.00, real=0.30 secs]
Heap after GC invocations=456 (full 0):
PSYoungGen      total 2091008K, used 1482K [0x0000000780000000, 0x0000000800000000, 0x0000000800000000)
  eden space 2084864K, 0% used [0x0000000780000000,0x0000000780000000,0x00000007ff400000)
  from space 6144K, 24% used [0x00000007ffa00000,0x00000007ffb72928,0x0000000800000000)
  to   space 6144K, 0% used [0x00000007ff400000,0x00000007ff400000,0x00000007ffa00000)
PSOldGen        total 3145728K, used 2267981K [0x00000006c0000000, 0x0000000780000000, 0x0000000780000000)
  object space 3145728K, 72% used [0x00000006c0000000,0x000000074a6d3528,0x0000000780000000)
PSPermGen       total 819200K, used 57100K [0x0000000680000000, 0x00000006b2000000, 0x00000006c0000000)
  object space 819200K, 6% used [0x0000000680000000,0x00000006837c3130,0x00000006b2000000)
}
{Heap before GC invocations=457 (full 0):
PSYoungGen      total 2091008K, used 2086346K [0x0000000780000000, 0x0000000800000000, 0x0000000800000000)
  eden space 2084864K, 100% used [0x0000000780000000,0x00000007ff400000,0x00000007ff400000)
  from space 6144K, 24% used [0x00000007ffa00000,0x00000007ffb72928,0x0000000800000000)
  to   space 6144K, 0% used [0x00000007ff400000,0x00000007ff400000,0x00000007ffa00000)
PSOldGen        total 3145728K, used 2267981K [0x00000006c0000000, 0x0000000780000000, 0x0000000780000000)
  object space 3145728K, 72% used [0x00000006c0000000,0x000000074a6d3528,0x0000000780000000)
PSPermGen       total 819200K, used 57100K [0x0000000680000000, 0x00000006b2000000, 0x00000006c0000000)
  object space 819200K, 6% used [0x0000000680000000,0x00000006837c3130,0x00000006b2000000)
1463.560: [GC
Desired survivor size 6291456 bytes, new threshold 1 (max 15)
[PSYoungGen: 2086346K->3177K(2091008K)] 4354327K->2271183K(5236736K), 0.2974600 secs] [Times: user=0.33 sys=0.00, real=0.30 secs]
Heap after GC invocations=457 (full 0):
PSYoungGen      total 2091008K, used 3177K [0x0000000780000000, 0x0000000800000000, 0x0000000800000000)
  eden space 2084864K, 0% used [0x0000000780000000,0x0000000780000000,0x00000007ff400000)
  from space 6144K, 51% used [0x00000007ff400000,0x00000007ff71a730,0x00000007ffa00000)
  to   space 6144K, 0% used [0x00000007ffa00000,0x00000007ffa00000,0x0000000800000000)
PSOldGen        total 3145728K, used 2268005K [0x00000006c0000000, 0x0000000780000000, 0x0000000780000000)
  object space 3145728K, 72% used [0x00000006c0000000,0x000000074a6d9528,0x0000000780000000)
PSPermGen       total 819200K, used 57100K [0x0000000680000000, 0x00000006b2000000, 0x00000006c0000000)
  object space 819200K, 6% used [0x0000000680000000,0x00000006837c3130,0x00000006b2000000)
}
{Heap before GC invocations=458 (full 0):
PSYoungGen      total 2091008K, used 2088041K [0x0000000780000000, 0x0000000800000000, 0x0000000800000000)
  eden space 2084864K, 100% used [0x0000000780000000,0x00000007ff400000,0x00000007ff400000)
  from space 6144K, 51% used [0x00000007ff400000,0x00000007ff71a730,0x00000007ffa00000)
  to   space 6144K, 0% used [0x00000007ffa00000,0x00000007ffa00000,0x0000000800000000)
PSOldGen        total 3145728K, used 2268005K [0x00000006c0000000, 0x0000000780000000, 0x0000000780000000)
  object space 3145728K, 72% used [0x00000006c0000000,0x000000074a6d9528,0x0000000780000000)
PSPermGen       total 819200K, used 57100K [0x0000000680000000, 0x00000006b2000000, 0x00000006c0000000)
  object space 819200K, 6% used [0x0000000680000000,0x00000006837c31b0,0x00000006b2000000)
1465.998: [GC
Desired survivor size 6291456 bytes, new threshold 1 (max 15)
[PSYoungGen: 2088041K->1456K(2091008K)] 4356047K->2269485K(5236736K), 0.2940590 secs] [Times: user=0.33 sys=0.00, real=0.29 secs]
Heap after GC invocations=458 (full 0):
PSYoungGen      total 2091008K, used 1456K [0x0000000780000000, 0x0000000800000000, 0x0000000800000000)
  eden space 2084864K, 0% used [0x0000000780000000,0x0000000780000000,0x00000007ff400000)
  from space 6144K, 23% used [0x00000007ffa00000,0x00000007ffb6c188,0x0000000800000000)
  to   space 6144K, 0% used [0x00000007ff400000,0x00000007ff400000,0x00000007ffa00000)
PSOldGen        total 3145728K, used 2268029K [0x00000006c0000000, 0x0000000780000000, 0x0000000780000000)
  object space 3145728K, 72% used [0x00000006c0000000,0x000000074a6df528,0x0000000780000000)
PSPermGen       total 819200K, used 57100K [0x0000000680000000, 0x00000006b2000000, 0x00000006c0000000)
  object space 819200K, 6% used [0x0000000680000000,0x00000006837c31b0,0x00000006b2000000)
}

CONFIDENTIALITY NOTICE This message and any included attachments are from Cerner Corporation and are intended only for the addressee. The information contained in this message is confidential and may constitute inside or non-public information under international, federal, or state securities laws. Unauthorized forwarding, printing, copying, distribution, or use of such information is strictly prohibited and may be unlawful. If you are not the addressee, please promptly delete this message and notify the sender of the delivery error by e-mail or you may call Cerner's corporate offices in Kansas City, Missouri, U.S.A at (+1) (816)221-1024.

_______________________________________________

hotspot-gc-use mailing list

hotspot-gc-use at openjdk.java.net<mailto:hotspot-gc-use at openjdk.java.net>

http://mail.openjdk.java.net/mailman/listinfo/hotspot-gc-use<https://urldefense.proofpoint.com/v2/url?u=http-3A__mail.openjdk.java.net_mailman_listinfo_hotspot-2Dgc-2Duse&d=CwMF-g&c=NRtzTzKNaCCmhN_9N2YJR-XrNU1huIgYP99yDsEzaJo&r=boyF4rmil1TfZ0CMFcwpJ5XChHGvDUZdECQt6wqy-r0&m=B6u7i_01oloEDuUMBMJsz1kisWSAdfbraatfZ8rVGDs&s=PhJEuNeIJsg-CJfmrUwLpxmlCBYo66v8VqK0denTqh8&e=>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/hotspot-gc-use/attachments/20160606/e44327d0/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Garbage.java
Type: application/octet-stream
Size: 2157 bytes
Desc: Garbage.java
URL: <http://mail.openjdk.java.net/pipermail/hotspot-gc-use/attachments/20160606/e44327d0/Garbage-0001.java>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: run.sh
Type: application/octet-stream
Size: 817 bytes
Desc: run.sh
URL: <http://mail.openjdk.java.net/pipermail/hotspot-gc-use/attachments/20160606/e44327d0/run-0001.sh>