cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Stu Hood (JIRA)" <j...@apache.org>
Subject [jira] Commented: (CASSANDRA-1488) Allow index sampling ratio to be configured in storage-conf
Date Fri, 17 Sep 2010 23:57:33 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-1488?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12910857#action_12910857
] 

Stu Hood commented on CASSANDRA-1488:
-------------------------------------

It seems like one of the only reasons to increase INDEX_INTERVAL would be to lower memory
usage. Decreasing it doesn't seem like a good idea, as you would want to increase the key
cache size instead. Would it make more sense for INDEX_INTERVAL to become a calculated value
based on a hard cap on memory usage?

> Allow index sampling ratio to be configured in storage-conf
> -----------------------------------------------------------
>
>                 Key: CASSANDRA-1488
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-1488
>             Project: Cassandra
>          Issue Type: New Feature
>          Components: Core
>            Reporter: Jonathan Ellis
>            Assignee: Jon Hermes
>            Priority: Minor
>             Fix For: 0.6.6
>
>         Attachments: 1488.txt
>
>
> Cassandra keeps a sample of row keys in memory to look up the right block of index entries,
to look up row data with.  This is kept in the SSTableReader.indexSummary field.  (See Google's
BigTable paper for a high-level description of how this index works; cassandra's is slightly
different in the details but the idea is the same.)
> We'd like to make IndexSummary.INDEX_INTERVAL configurable in storage-conf.xml, i.e.,
in the DatabaseDescriptor class.  For 0.6, a global setting is fine.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message