You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@cassandra.apache.org by "Jonathan Ellis (JIRA)" <ji...@apache.org> on 2010/07/21 01:11:50 UTC
[jira] Commented: (CASSANDRA-1302) Allow Row Iterator to use the
RowCache
[ https://issues.apache.org/jira/browse/CASSANDRA-1302?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12890494#action_12890494 ]
Jonathan Ellis commented on CASSANDRA-1302:
-------------------------------------------
(a) we don't want to pollute the row cache w/ data from range queries, since you can completely displace your hot rows w/ a single RQ of rows that will never be used again.
(b) the reason RIF is going through so much drama there is we are going through a lot of effort to be able to re-use the readers and filters from one row on the next one. once you're scanning through non-hot data (e.g. most hadoop jobs) using normal getCF is going to be much slower.
(c) it should be okay to check the row cache as a fast-path though. maybe even populate the row cache _if_ the row cache is not full to capacity.
> Allow Row Iterator to use the RowCache
> --------------------------------------
>
> Key: CASSANDRA-1302
> URL: https://issues.apache.org/jira/browse/CASSANDRA-1302
> Project: Cassandra
> Issue Type: Improvement
> Components: Core
> Reporter: T Jake Luciani
> Attachments: row-iterator-cache-patch.txt
>
>
> Range slices are very slow.
> I've discovered this is caused by the RowIterator ignoring the row cache.
> I've altered the code to use the row cache and now see a factor of 30 performance boost.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.