11/24/2023 0 Comments Beats updater resource hog![]() In our case, we saw that C4 reduces latency (P90 ~17ms) compared with CMS (P90 ~24ms). Next generation GC algorithms, such as C4, ZGC, and Shenandoah, show promising results. Traditional GC algorithms have trouble handling this scale and experience long GC pause times. In a large scale data infrastructure, critical services usually have very large JVM heap sizes, ranging from several hundreds of gigabytes to terabytes. In tuning Presto GC, we found the String Deduplication setting dominates the performance impact. For example, we found changing Young Generation size and ParGCCardsPerStrideChunk improved the performance significantly, but we did not see much difference when changing TLABSize and ConcGCThreads. There are many GC parameters to tune, but usually changing one or two parameters makes significant impact while others are negligible.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |