[
https://issues.apache.org/jira/browse/HBASE-20542?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16529293#comment-16529293
]
Duo Zhang commented on HBASE-20542:
-----------------------------------
[~stack] has been doing the performance testing for 2.0.x, there is no much
difference between 2.1.x and 2.0.x I believe, most performance related patches
are applied to both branches.
Turn on IMC by default maybe too late for 2.1.0, let's do it in 2.2? We have
also planned to fix the AMv2 on branch-2, see HBASE-20828, so 2.1 will not be
the last minor release line of 2.x. Take it easy.
> Better heap utilization for IMC with MSLABs
> -------------------------------------------
>
> Key: HBASE-20542
> URL: https://issues.apache.org/jira/browse/HBASE-20542
> Project: HBase
> Issue Type: Sub-task
> Reporter: Eshcar Hillel
> Assignee: Eshcar Hillel
> Priority: Major
> Attachments: HBASE-20542.branch-2.001.patch,
> HBASE-20542.branch-2.003.patch, HBASE-20542.branch-2.004.patch,
> HBASE-20542.branch-2.005.patch, HBASE-20542.master.003.patch, run.sh,
> workloada, workloadc, workloadx, workloady
>
>
> Following HBASE-20188 we realized in-memory compaction combined with MSLABs
> may suffer from heap under-utilization due to internal fragmentation. This
> jira presents a solution to circumvent this problem. The main idea is to have
> each update operation check if it will cause overflow in the active segment
> *before* it is writing the new value (instead of checking the size after the
> write is completed), and if it is then the active segment is atomically
> swapped with a new empty segment, and is pushed (full-yet-not-overflowed) to
> the compaction pipeline. Later on the IMC deamon will run its compaction
> operation (flatten index/merge indices/data compaction) in the background.
> Some subtle concurrency issues should be handled with care. We next elaborate
> on them.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
