You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hive.apache.org by "Rajesh Balamohan (JIRA)" <ji...@apache.org> on 2017/05/24 04:37:04 UTC

[jira] [Updated] (HIVE-16746) Reduce number of index lookups for same table in IndexWhereTaskDispatcher

     [ https://issues.apache.org/jira/browse/HIVE-16746?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Rajesh Balamohan updated HIVE-16746:
------------------------------------
    Attachment: HIVE-16746.1.patch

With patch, response time came down from 185 seconds to 140-145 seconds. This benefits irrespective of store implementation. Currently CachedStore does not yet implement index caching.

> Reduce number of index lookups for same table in IndexWhereTaskDispatcher
> -------------------------------------------------------------------------
>
>                 Key: HIVE-16746
>                 URL: https://issues.apache.org/jira/browse/HIVE-16746
>             Project: Hive
>          Issue Type: Bug
>            Reporter: Rajesh Balamohan
>            Priority: Minor
>         Attachments: HIVE-16746.1.patch
>
>
> {{IndexWhereTaskDispatcher}} is used when {{hive.optimize.index.filter=true}}. It lists all indices for the table and depending on the query complexity, this ends up being in the hotpath. For e.g, Q14 explain plan takes 180-200 seconds and this index querying multiple times for same tables take up 30-40 seconds. This function was invoked around 24000 times for same set of tables.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)