[
https://issues.apache.org/jira/browse/CASSANDRA-325?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12740813#action_12740813
]
Jonathan Ellis commented on CASSANDRA-325:
------------------------------------------
Having thought it over I think we should keep the bloom filter and activate it
as in the -v2 patch.
The bloom filter will be more useful than block indexes alone in the case of
doing a name slice, which is the common case for most apps.
Additionally, FB appears to have been moving towards the filters rather than
away -- the code uploaded to code.google.com didn't have them at all, and the
original apache upload did. (One of the read paths checked it; in another it
was commented out. That it disappeared entirely as we cleaned things up is
probably my fault.)
> Column bloomfilter not exploited
> --------------------------------
>
> Key: CASSANDRA-325
> URL: https://issues.apache.org/jira/browse/CASSANDRA-325
> Project: Cassandra
> Issue Type: Improvement
> Components: Core
> Reporter: Jun Rao
> Assignee: Jonathan Ellis
> Fix For: 0.4
>
> Attachments: 325-v2.patch, 325.patch
>
>
> There is a bloomfilter created for all columns per row in SSTables. The
> bloomfilter is deserialized during read, but is not checked.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
