You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@kudu.apache.org by "Todd Lipcon (JIRA)" <ji...@apache.org> on 2016/05/05 23:07:12 UTC

[jira] [Commented] (KUDU-815) Improve performance of first scan following TS restart

    [ https://issues.apache.org/jira/browse/KUDU-815?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15273285#comment-15273285 ] 

Todd Lipcon commented on KUDU-815:
----------------------------------

Looking at some stress tests via Impala on a cluster, and want to pick up this and a couple related optimizations.

> Improve performance of first scan following TS restart
> ------------------------------------------------------
>
>                 Key: KUDU-815
>                 URL: https://issues.apache.org/jira/browse/KUDU-815
>             Project: Kudu
>          Issue Type: Bug
>          Components: perf, tablet
>    Affects Versions: Private Beta
>            Reporter: Adar Dembo
>            Assignee: Todd Lipcon
>            Priority: Minor
>
> Because cfiles are opened lazily, the very first scan following a TS restart cannot cull the set of scanned delta files based on minimum/maximum timestamp (the min/max are found in the file's delta stats, which aren't loaded until the file is opened). This makes the first scan quite slow.
> We should find some way to work around this to improve the scan's performance.



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