You are viewing a plain text version of this content. The canonical link for it is here.

Posted to dev@hive.apache.org by "Xuefu Zhang (JIRA)" <ji...@apache.org> on 2014/09/22 22:19:36 UTC

[jira] [Created] (HIVE-8219) Multi-Insert optimization, don't sink the source into a file [Spark Branch]

Xuefu Zhang created HIVE-8219:
---------------------------------

             Summary: Multi-Insert optimization, don't sink the source into a file [Spark Branch]
                 Key: HIVE-8219
                 URL: https://issues.apache.org/jira/browse/HIVE-8219
             Project: Hive
          Issue Type: Bug
          Components: Spark
            Reporter: Xuefu Zhang


Current implementation split the operator plan at the lowest common ancester by inserting one FileSinkOperator and a list of TableScanOperators. Writing to a file (by the FS) is expensive. We should be able to insert a ReduceSinkOperator instead. The result RDD from the first job can be cached and refereed in subsequent Spark jobs.

This is a followup for HIVE-7503.



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