[ https://issues.apache.org/jira/browse/CASSANDRA-9754?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15565823#comment-15565823 ]
Michael Kjellman commented on CASSANDRA-9754: --------------------------------------------- Stable all night! My large test partitions have grown to ~12.5GB Just as stable -- latencies are unchanged. I'm so happy!!! ~7ms avg p99.9th and ~925 microseconds average read latency. GC basically non-existant -- and for what GC is happening, the instances are averaging a 111 microsecond ParNew collection -- almost NO CMS! Compaction is keeping up. On the converse side (the control 2.1 cluster running the same load) has instances are OOMing left and right -- CMS is frequently running 250 ms collections, ParNew is running 1.28 times a second on average with 75 ms average ParNew times. Horrible! And that's average -- the upper percentiles are a mess so I won't bore everyone. Read latencies are currently 380 ms average with many 15 *second* read latencies in the p99.9. > Make index info heap friendly for large CQL partitions > ------------------------------------------------------ > > Key: CASSANDRA-9754 > URL: https://issues.apache.org/jira/browse/CASSANDRA-9754 > Project: Cassandra > Issue Type: Improvement > Reporter: sankalp kohli > Assignee: Michael Kjellman > Priority: Minor > Fix For: 4.x > > Attachments: 9754_part1-v1.diff, 9754_part2-v1.diff > > > Looking at a heap dump of 2.0 cluster, I found that majority of the objects > are IndexInfo and its ByteBuffers. This is specially bad in endpoints with > large CQL partitions. If a CQL partition is say 6,4GB, it will have 100K > IndexInfo objects and 200K ByteBuffers. This will create a lot of churn for > GC. Can this be improved by not creating so many objects? -- This message was sent by Atlassian JIRA (v6.3.4#6332)