On Wed, 28 Jun 2023 08:04:21 GMT, Yi Yang <yy...@openjdk.org> wrote: >> ### Motivation and proposal >> Hi, heap dump brings about pauses for application's execution(STW), this is >> a well-known pain. JDK-8252842 have added parallel support to heapdump in an >> attempt to alleviate this issue. However, all concurrent threads >> competitively write heap data to the same file, and more memory is required >> to maintain the concurrent buffer queue. In experiments, we did not feel a >> significant performance improvement from that. >> >> The minor-pause solution, which is presented in this PR, is a two-stage >> segmented heap dump: >> >> 1. Stage One(STW): Concurrent threads directly write data to multiple heap >> files. >> 2. Stage Two(Non-STW): Merge multiple heap files into one complete heap dump >> file. >> >> Now concurrent worker threads are not required to maintain a buffer queue, >> which would result in more memory overhead, nor do they need to compete for >> locks. The changes in the overall design are as follows: >> >>  >> <p align="center">Fig1. Before</p> >> >>  >> <p align="center">Fig2. After this patch</p> >> >> ### Performance evaluation >> | memory | numOfThread | STW | Total | Compression | >> | --- | --------- | -------------- | ------------ | -------- | >> | 8g | 1 thread | 15.612 secs | 15.612 secs | N | >> | 8g | 32 thread | 2.5617250 secs | 14.498 secs | N | >> | 8g | 32 thread | 2.3084878 secs | 3.198 secs | Compress1 | >> | 8g | 32 thread | 10.9355128 secs | 21.882 secs | Compress2 | >> | 8g | 96 thread | 2.6790452 secs | 14.012 secs | N | >> | 8g | 96 thread | 2.3044796 secs | 3.589 secs | Compress1 | >> | 8g | 96 thread | 9.7585151 secs | 20.219 secs| Compress2 | >> | 16g | 1 thread | 26.278 secs | 26.278 secs | N | >> | 16g | 32 thread | 5.2313740 secs | 26.417 secs | N | >> | 16g | 32 thread | 5.6946983 secs | 6.538 secs | Compress1 | >> | 16g | 32 thread | 21.8211105 secs | 41.133 secs | Compress2 | >> | 16g | 96 thread | 6.2445556 secs | 27.141 secs | N | >> | 16g | 96 thread | 4.6007096 secs | 6.259 secs | Compress1 | >> | 16g | 96 thread | 19.2965783 secs | 39.007 secs | Compress2 | >> | 32g | 1 thread | 48.149 secs | 48.149 secs | N | >> | 32g | 32 thread | 10.7734677 secs | 61.643 secs | N | >> | 32g | 32 thread | 10.1642097 secs | 10.903 secs | Compress1 | >> | 32g | 32 thread | 43.8407607 secs | 88.152 secs | Compress2 | >> | 32g | 96 thread | 13.1522042 ... > > Yi Yang has updated the pull request incrementally with one additional commit > since the last revision: > > memory leak
I think commit "use HandshakeClosure instead of VMOperation" is a wrong way to go. It restricts use of parallel dumping only to attach case. I have pending changes in heap dumper to support virtual threads and I'm going switch heap dumper to always use DumpMerger. ------------- PR Comment: https://git.openjdk.org/jdk/pull/13667#issuecomment-1615243585
