[
https://issues.apache.org/jira/browse/HIVE-10155?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14553295#comment-14553295
]
Sergey Shelukhin commented on HIVE-10155:
-----------------------------------------
Will look at cache behavior. Perf wise LRFU is not slower
> LLAP: switch to sensible cache policy
> -------------------------------------
>
> Key: HIVE-10155
> URL: https://issues.apache.org/jira/browse/HIVE-10155
> Project: Hive
> Issue Type: Sub-task
> Reporter: Sergey Shelukhin
> Assignee: Sergey Shelukhin
>
> FIFO policy is currently the default. We can test LRFU one but there were
> concerns that it won't scale. One option is to implement a two-tier policy,
> FIFO/LRU for blocks referenced once, and something more complex (LFU, LRFU)
> for blocks referenced more than once. That should be friendly to large scans
> of a fact table in terms of behavior and overhead.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
