cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Carl Yeksigian (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-9830) Option to disable bloom filter in highest level of LCS sstables
Date Tue, 04 Oct 2016 15:55:21 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-9830?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15545808#comment-15545808
] 

Carl Yeksigian commented on CASSANDRA-9830:
-------------------------------------------

I'm +1 on the code changes to the compaction strategy, but the perf testing never ran, so
we haven't yet shown a reduction in memory usage - [~pauloricardomg]: can you take a look
and ensure it runs (or post the results if they have run properly).

Also, I'd like to see some additional tests that make sure that we are getting the same set
of results whether or not the bloom filter is loaded for the top level. Should be both point
queries and range queries.

> Option to disable bloom filter in highest level of LCS sstables
> ---------------------------------------------------------------
>
>                 Key: CASSANDRA-9830
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-9830
>             Project: Cassandra
>          Issue Type: New Feature
>          Components: Compaction
>            Reporter: Jonathan Ellis
>            Assignee: Paulo Motta
>            Priority: Minor
>              Labels: lcs, performance
>             Fix For: 3.x
>
>
> We expect about 90% of data to be in the highest level of LCS in a fully populated series.
 (See also CASSANDRA-9829.)
> Thus if the user is primarily asking for data (partitions) that has actually been inserted,
the bloom filter on the highest level only helps reject sstables about 10% of the time.
> We should add an option that suppresses bloom filter creation on top-level sstables.
 This will dramatically reduce memory usage for LCS and may even improve performance as we
no longer check a low-value filter.
> (This is also an idea from RocksDB.)



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

Mime
View raw message