hive-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sergey Shelukhin (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HIVE-11355) Hive on tez: memory manager for sort buffers (input/output) and operators
Date Fri, 07 Aug 2015 19:49:46 GMT

    [ https://issues.apache.org/jira/browse/HIVE-11355?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14662329#comment-14662329
] 

Sergey Shelukhin commented on HIVE-11355:
-----------------------------------------

Some comments in evaluateUnionWork? Input size is computed from a subset of output edges,
but no input edges are taken into account
Comment for "forAll". Actually I don't understand the difference between using regex rule
for mapjoins and the default rule that checks for mapjoins. Aren't they supposed to do the
same thing?
{noformat}
           throw new SemanticException("Memory shortage of " + -(totalAvailableMemory)
                + ". Please modify the container size to be greater than "
                + HiveConf.ConfVars.HIVECONVERTJOINNOCONDITIONALTASKTHRESHOLD);
{noformat}
isn't it supposed to already be addressed by the above exception? At least, the message seems
to indicate so.
pctx.conf.getLongVar(HiveConf.ConfVars.MAPREDMAXSPLITSIZE - can this be 0?
if (edge.getDataFlowSize() > TEN_MB) { nit - indentation.
To be continued
Actually RB would help :)


> Hive on tez: memory manager for sort buffers (input/output) and operators
> -------------------------------------------------------------------------
>
>                 Key: HIVE-11355
>                 URL: https://issues.apache.org/jira/browse/HIVE-11355
>             Project: Hive
>          Issue Type: Improvement
>          Components: Tez
>    Affects Versions: 2.0.0
>            Reporter: Vikram Dixit K
>            Assignee: Vikram Dixit K
>         Attachments: HIVE-11355.1.patch, HIVE-11355.2.patch, HIVE-11355.3.patch
>
>
> We need to better manage the sort buffer allocations to ensure better performance. Also,
we need to provide configurations to certain operators to stay within memory limits.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message