Java:Increasing YoungGen size to improve GC performance

Question

I was reading the following article: http://java.sun.com/docs/hotspot/gc1.4.2/example.html and have trouble understanding the following lines:

Young generation size is too small

The young generation heap size in this first example is about 4 Mbytes with an overall heap size of about 32 Mbytes.

[GC [DefNew: 4032K->64K(4032K), 0.0429742 secs] 9350K->7748K(32704K), 0.0431096 secs]

[GC [DefNew: 4032K->64K(4032K), 0.0403446 secs] 11716K->10121K(32704K), 0.0404867 secs]

[GC [DefNew: 4032K->64K(4032K), 0.0443969 secs] 14089K->12562K(32704K), 0.0445251 secs]

This output seems reasonable from the point of view of the time spent in garbage collection but note that although the occupancy of the young generation is decreasing (e.g., in the first line from 4032 to 64k by 3968k) the occupancy of the entire heap is decreasing by considerably less (by 1602k from 9350k to 7748k). This indicates that only about 40% objects in the young generation were garbage and the rest survive the collection and are being promoted into the old generation.


Increasing the young generation size to increase the free space after the collection


The young generation heap size in this next run is increased to 8 Mbytes.


[GC [DefNew: 8128K->64K(8128K), 0.0453670 secs] 13000K->7427K(32704K), 0.0454906 secs]

[GC [DefNew: 8128K->64K(8128K), 0.0388632 secs] 15491K->9663K(32704K), 0.0390013 secs]

[GC [DefNew: 8128K->64K(8128K), 0.0388610 secs] 17727K->11829K(32704K), 0.0389919 secs]


With an 8 Mbyte size most of young generation is garbage at the time of the minor collection. In the first line the young generation heap decreases by 8064k from 8128k to 64k and the entire heap decreases by 5573k from 13000k to 7427k meaning about 68% of the young generation was garbage. As would be expected the size of the young generation does not change the amount of live objects that survive the minor collection (about 2.4M bytes in each case) but there are fewer minor collections and the cost of the collections in terms of the minor collection pause times are comparable (fractions of a second listed at the end of each line).

My question is, how does increasing the size of YoungGen helps us in this case. The total number of objects that application allocates in YoungGen will be constant.

Answer 1

The answer lies in the article itself.

As would be expected the size of the young generation does not change the amount of live objects that survive the minor collection (about 2.4M bytes in each case) but there are fewer minor collections and the cost of the collections in terms of the minor collection pause times are comparable (fractions of a second listed at the end of each line).

This is a direct consequence of increasing the YoungGen.

Answer 2

Promoting something out of the YoungGen is expensive, it causes:

• it to live for longer, wasting memory
• future garbage collections (every one until it's dead) to be more expensive
• it to be ignored by YoungGen collections, and you having to wait until you exhaust a more precious resource, and force a more expensive collection.

By increasing the size of the YoungGen, they are ensuring that more of the objects are collected by it, and hence not hitting any of the bad points above. That is, the object dies quickly and doesn't cost anyone anything bad, so everything is faster?

Additionally, there's a much more modern version of the document you're reading, although it seems to lack your example: http://www.oracle.com/technetwork/java/javase/gc-tuning-6-140523.html

Answer 3

The cost of a young collection is proportional to the number of objects retained. The larger the young collection the more likely that those objects are no longer required, meaning you can find that the cost of a young collection doesn't increase much beyond a certain size.

For small young gen sizes its proportional but for larger sizes not so much. The frequency of collection drops proportionally with size. You can reach the point that its collecting less than once per day. ;)

YMMV depending on how your application behaves.