Todd Lipcon has posted comments on this change. Change subject: process_memory: go back to non-incremental tracking ......................................................................
Patch Set 1: (2 comments) http://gerrit.cloudera.org:8080/#/c/6915/1//COMMIT_MSG Commit Message: PS1, Line 34: noticeably better > Thanks. It looks like the difference between the two implementations is abo I think it's more significant -- eg, where the red line crosses 20 us/row, the blue line is around 18.5 us/row. So, that's 7.5% or so. http://gerrit.cloudera.org:8080/#/c/6915/1/src/kudu/util/process_memory.cc File src/kudu/util/process_memory.cc: Line 191: const int64_t kReadIntervalMicros = 50000; > I thought we could consume more than that, even in such a short period of t If we're doing 1M inserts/sec, then it must be with a smaller row size (1M * 1k row size = 1GB/sec, which is more than I've seen us ingest per node). And I think in that case, overshooting by 50M isn't so bad, either, considering to sustain 1GB ingest per second you must have a relatively high memory limit, lots of maintenance threads, etc, and thus 50M is only a tiny percentage of your overall heap, right? -- To view, visit http://gerrit.cloudera.org:8080/6915 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: I8823028de3ea260f1450d9bf34af2dc5a794b206 Gerrit-PatchSet: 1 Gerrit-Project: kudu Gerrit-Branch: master Gerrit-Owner: Todd Lipcon <t...@apache.org> Gerrit-Reviewer: Adar Dembo <a...@cloudera.com> Gerrit-Reviewer: Kudu Jenkins Gerrit-Reviewer: Todd Lipcon <t...@apache.org> Gerrit-HasComments: Yes