hive-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Xuefu Zhang (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HIVE-8262) Create CacheTran that transforms the input RDD by caching it [Spark Branch]
Date Tue, 30 Sep 2014 18:42:35 GMT

    [ https://issues.apache.org/jira/browse/HIVE-8262?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14153537#comment-14153537
] 

Xuefu Zhang commented on HIVE-8262:
-----------------------------------

Let's put this one on hold until we find out if it's simpler just to put a caching flag in
other SparkTran subclasses. 

> Create CacheTran that transforms the input RDD by caching it [Spark Branch]
> ---------------------------------------------------------------------------
>
>                 Key: HIVE-8262
>                 URL: https://issues.apache.org/jira/browse/HIVE-8262
>             Project: Hive
>          Issue Type: Sub-task
>          Components: Spark
>            Reporter: Xuefu Zhang
>            Assignee: Chao
>         Attachments: HIVE-8262.1-spark.patch
>
>
> In a few cases we need to cache a RDD to avoid recompute it for better performance. However,
caching a map input RDD is different from caching a regular RDD due to SPARK-3693. The way
to cache a Hadoop RDD, which is the input to MapWork, is to cache, the result RDD that is
transformed from the original Hadoop RDD by applying a map function, in which <key, value>
pairs are copied. To cache intermediate RDDs, such as that from a shuffle, is just calling
.cache().
> This task is to create a CacheTran to capture this, which can be used to plug in Spark
Plan when caching is desirable. 



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

Mime
View raw message