hive-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Siddharth Seth (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HIVE-12470) Allow splits to provide custom consistent locations, instead of being tied to data locality
Date Thu, 17 Dec 2015 03:20:46 GMT

    [ https://issues.apache.org/jira/browse/HIVE-12470?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15061390#comment-15061390
] 

Siddharth Seth commented on HIVE-12470:
---------------------------------------

RB already exists.
On the sorting - the list can change on each refresh, and it isn't known whether the list
actually changes or not. That could be tracked. However, given this is not accessed very frequently,
I did not try to optimize away the sort.
Cache registries by name - for a single client which may communicate with different llap instances.
e.g. a single hive server instance which can submit to different llap daemons.

> Allow splits to provide custom consistent locations, instead of being tied to data locality
> -------------------------------------------------------------------------------------------
>
>                 Key: HIVE-12470
>                 URL: https://issues.apache.org/jira/browse/HIVE-12470
>             Project: Hive
>          Issue Type: Improvement
>          Components: llap
>            Reporter: Siddharth Seth
>            Assignee: Siddharth Seth
>         Attachments: HIVE-12470.1.txt, HIVE-12470.1.wip.txt
>
>
> LLAP instances may not run on the same nodes as HDFS, or may run on a subset of the cluster.
> Using split locations based on FileSystem locality is not very useful in such cases -
since that guarantees not getting any locality.
> Allow a split to map to a specific location - so that there's a chance of getting cache
locality across different queries.



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

Mime
View raw message