[
https://issues.apache.org/jira/browse/HIVE-20250?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16662571#comment-16662571
]
Thai Bui commented on HIVE-20250:
---------------------------------
Hi [~jdere], I'm talking a look at this ticket now and would like to offer some
help to accelerate the process. I would like to take on the task with your
guidance if possible.
My company and my team is planing to accelerate our BI tools and having Hive as
an options to work with external tables and query result cache would greatly
benefit us. Thus this request.
>From the previous conversation, it seems having a result timeout is the
>simplest approach, however, I am thinking of an additional approach (supported
>in conjunction with the timeouts) where HS2 and its external tables mechanism
>would receive notifications to invalidate the result cache. This is
>approachable because most of the blob storage systems (such as S3) have
>delete/modify/etc. change events that we could subscribe to.
This is a completely different way of thinking about managed/external tables in
Hive. I think we can evenly marry the two and there's no boundary between what
is a managed vs. what is an external tables anymore. I'm not sure how feasible
this is, but if managed/transactional tables and their APIs are well-defined,
they could be refactored into actors in an reactive model to listen to
messages, where each message could be to invalidate a cache, etc.
Let me know what you think. The first step could be just to work on the timeout
approach since it is relatively simple.
> Option to allow external tables to use query results cache
> ----------------------------------------------------------
>
> Key: HIVE-20250
> URL: https://issues.apache.org/jira/browse/HIVE-20250
> Project: Hive
> Issue Type: Sub-task
> Reporter: Jason Dere
> Priority: Major
> Attachments: HIVE-20250.1.patch
>
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
