JEP 291: Deprecate the Concurrent Mark Sweep (CMS) Garbage Collector

[Thomas Schatzl]

Hi again,

On Sun, 2017-04-16 at 18:26 +0200, Thomas Schatzl wrote:
>
> Please see the JavaOne G1 presentation for some more details [3].
[...]
> [3] https://oracle.rainfocus.com/scripts/catalog/oow16.jsp?event=java
> one&search=CON3889&search.event=javaone

  of course I got the reference wrong, the one I meant is CON3815 [1].
Apologies.

And in case you are curious about the scope of improvements of G1 JDK 9
vs. 8u on huge workloads, have a look at CON1516: "Java 9: The Quest
for Very Large Heaps" [2].

This is an example with requirements and setup way beyond the typical
application (we can call it "big toy-application" for this discussion),
so its applicability to your workloads is of course rather questionable
(I only mention it because when looking up the G1 JDK 9 presentation
for JavaOne this one also showed up, I agree it is way too different to
yours). However all of the changes to make this possible also change G1
behavior on much smaller heaps on much smaller machines significantly -
but you have to trust me here that these changes are desirable :)

As for testing, I think the last significant commit related to G1
performance for JDK 9 has been pushed to master mid-September last year
[3], most of them months earlier. This means there has been opportunity
for running JDK 9 on any (non-Oracle) actual production workloads with
basically the final version of G1 by the community for seven months now
for anyone, and contribute their experience.

Thanks,
  Thomas

[1] https://oracle.rainfocus.com/scripts/catalog/oow16.jsp?event=javaon
e&search=CON3815&search.event=javaone
[2] https://oracle.rainfocus.com/scripts/catalog/oow16.jsp?event=javaon
e&search=CON1516&search.event=javaone
[3] I think https://bugs.openjdk.java.net/browse/JDK-8159422