[ https://issues.apache.org/jira/browse/CASSANDRA-6890?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13984805#comment-13984805 ]
Pavel Yaskevich commented on CASSANDRA-6890: -------------------------------------------- Just mine 2 cents here. To archive those results did you use stress in the mixed mode and/or were reads happening simultaneously with writes. This regression could be caused by contention on the vm_area_struct access inside mm_struct. Can you do 'perf record' on the Cassandra process while stress is running and share the results? bq. It's somewhat interesting to note that buffered I/O uses roughly 35% more privileged cpu-time than memory-mapped, whereas user, interrupt, and all other processor counters show comparable load between the two configurations. This is expected because buffered i/o actually does syscall per read and has to copy data over from page cache to the user buffer. > Remove mmap'ed i/o from read path > --------------------------------- > > Key: CASSANDRA-6890 > URL: https://issues.apache.org/jira/browse/CASSANDRA-6890 > Project: Cassandra > Issue Type: Improvement > Reporter: Joshua McKenzie > Assignee: Joshua McKenzie > Priority: Minor > Fix For: 3.0 > > Attachments: mmap_gc.jpg, mmap_jstat.txt, mmap_perf.txt, > nommap_gc.jpg, nommap_jstat.txt > > > Since we actively unmap unreferenced SSTR's and also copy data out of those > readers on the read path, the current memory mapped i/o is a lot of > complexity for very little payoff. Clean out the mmapp'ed i/o on the read > path. -- This message was sent by Atlassian JIRA (v6.2#6252)