[
https://issues.apache.org/jira/browse/HIVE-3603?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13728202#comment-13728202
]
Swarnim Kulkarni commented on HIVE-3603:
----------------------------------------
[~navis] Any specific reason we chose not to provide a default value for
caching? In this case, it seems like the caching would only kick in if the user
explicitly demands for it by specifying the property "hbase.scan.cache" as a
part of the DDL. However I don't see any case where caching to a default value
won't be useful. Thoughts?
> Enable client-side caching for scans on HBase
> ---------------------------------------------
>
> Key: HIVE-3603
> URL: https://issues.apache.org/jira/browse/HIVE-3603
> Project: Hive
> Issue Type: Improvement
> Components: HBase Handler
> Reporter: Karthik Ranganathan
> Assignee: Navis
> Priority: Minor
> Fix For: 0.12.0
>
> Attachments: HIVE-3603.D7761.1.patch
>
>
> HBaseHandler sets up a TableInputFormat MR job against HBase to read data in.
> The underlying implementation (in HBaseHandler.java) makes an RPC call per
> row-key, which makes it very inefficient. Need to specify a client side cache
> size on the scan.
> Note that HBase currently only supports num-rows based caching (no way to
> specify a memory limit). Created HBASE-6770 to address this.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
