[
https://issues.apache.org/jira/browse/HBASE-16417?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15430054#comment-15430054
]
Anoop Sam John commented on HBASE-16417:
----------------------------------------
Mostly looks good. Though on the general use case (where not many
updates/deletes) why cannot we flush all the segments in pipeline together when
a flush to disk arise? In that case also, doing an in memory compaction for
segments in pipeline (eg: You say when segments# >3) is to reduce #files
flushed to disk. So another way for that is flush whole pipeline together. In
fact I feel at flush to file comes, we should be flushing all segments in
pipeline + active. So it is just like default memstore other than the in btw
flush to in memory flattened structure. When MSLAB in place, CellChunkMap
would be ideal. For off heap, any way we must need it, As a first step,
CellArrayMap being default is fine.
And good to see that ur tests reveal the overhead of scan for compact decision.
And ya we should be doing that with out any compaction based test. And ya it is
upto the user to know the pros and cons of in memory compaction and select that
wisely. We should be well documenting that..
Great.. We are mostly in sync now :-)
> In-Memory MemStore Policy for Flattening and Compactions
> --------------------------------------------------------
>
> Key: HBASE-16417
> URL: https://issues.apache.org/jira/browse/HBASE-16417
> Project: HBase
> Issue Type: Sub-task
> Reporter: Anastasia Braginsky
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
