[
https://issues.apache.org/jira/browse/HIVE-10154?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sergey Shelukhin resolved HIVE-10154.
-------------------------------------
Resolution: Fixed
Fix Version/s: llap
This adds an object pool, and pooling for the following objects:
ColumnVectorBatch (small pool per split, so that ColumnVector-s could be reused)
(the rest of the pools are global)
ColumnReadContext and ColumnReadContext.StreamContext (not sure these are
needed, they are always used within one method call, GC could be very efficient
for those)
OrcEncodedColumnBatch+OrcEncodedColumnBatch.StreamBuffer (passed between
threads/classes), TrackedCacheChunk/ProcCacheChunk (between methods, may also
be unnecessary).
> LLAP: GC issues 1
> -----------------
>
> Key: HIVE-10154
> URL: https://issues.apache.org/jira/browse/HIVE-10154
> Project: Hive
> Issue Type: Sub-task
> Reporter: Sergey Shelukhin
> Assignee: Sergey Shelukhin
> Fix For: llap
>
>
> Lots of small objects like CVBs, CVs, CacheChunk-s, BufferChunk-s, etc.
> floating around. Need to pool them, or reuse is a simpler way when possible
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
