[ https://issues.apache.org/jira/browse/HBASE-14613?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15207460#comment-15207460 ]
Elliott Clark commented on HBASE-14613: --------------------------------------- Yeah g1 no mslab. G1 really really loves having a balance of new and old gen. It holds off on a mixed gen collection as long as possible. So when we have a significant amount of memory basically pinned in old gen forever, it skews that balance pretty badly. That means that if you're promoting a decent amount of objects but not getting enough mixed gen collections, it's possible to get pauses even though there is a lot of memory left. To combat this we have just used more of the L2 cache offheap. This means that we can keep the young and old gens more balanced and don't risk having promotion stalls. > Remove MemStoreChunkPool? > ------------------------- > > Key: HBASE-14613 > URL: https://issues.apache.org/jira/browse/HBASE-14613 > Project: HBase > Issue Type: Brainstorming > Reporter: Lars Hofhansl > Priority: Minor > Attachments: 14613-0.98.txt, gc.png, writes.png > > > I just stumbled across MemStoreChunkPool. The idea behind is to reuse chunks > of allocations rather than letting the GC handle this. > Now, it's off by default, and it seems to me to be of dubious value. I'd > recommend just removing it. -- This message was sent by Atlassian JIRA (v6.3.4#6332)