[
https://issues.apache.org/jira/browse/CASSANDRA-4234?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Yuki Morishita updated CASSANDRA-4234:
--------------------------------------
Attachment: 4234.txt
(Attaching modified and rebased patch)
So, after running some benchmark, I concluded that getOverlappingSSTables is
not going to be a bottleneck. After all it's O(log n) operation.
But the problem rises if you iterate on 50K (or large number of) sstables. So I
modified the patch to first sort sstables by droppable ratio in descent order,
and skip iteration if it finds the ratio below threshold. I think this feature
combined with the nature of LCS (fewer overlap among sstables) prevent a lot of
calculation in findDroppableSSTable.
> Add tombstone-removal compaction to LCS
> ---------------------------------------
>
> Key: CASSANDRA-4234
> URL: https://issues.apache.org/jira/browse/CASSANDRA-4234
> Project: Cassandra
> Issue Type: Improvement
> Affects Versions: 1.2
> Reporter: Jonathan Ellis
> Assignee: Yuki Morishita
> Priority: Minor
> Labels: compaction
> Fix For: 1.2
>
> Attachments: 4234.txt
>
>
> CASSANDRA-3442 will recompact sstables with high levels of expired
> tombstones, but only under SCS.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
