cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "sankalp kohli (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-6474) Compaction strategy based on MinHash
Date Tue, 28 Jan 2014 15:43:38 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-6474?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13884260#comment-13884260
] 

sankalp kohli commented on CASSANDRA-6474:
------------------------------------------

I like the first approach since it is simpler in which we will have HLL using 1K per sstable.

 
Regarding you second approach, what do you mean by "For LCS it would only be useful for prioritization"?
So you will calculate the stable set based on description in  CASSANDRA-6216 but I am not
sure how you will use HLL for prioritization. 

> Compaction strategy based on MinHash
> ------------------------------------
>
>                 Key: CASSANDRA-6474
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-6474
>             Project: Cassandra
>          Issue Type: New Feature
>          Components: Core
>            Reporter: Yuki Morishita
>            Assignee: sankalp kohli
>              Labels: compaction
>             Fix For: 3.0
>
>
> We can consider an SSTable as a set of partition keys, and 'compaction' as de-duplication
of those partition keys.
> We want to find compaction candidates from SSTables that have as many same keys as possible.
If we can group similar SSTables based on some measurement, we can achieve more efficient
compaction.
> One such measurement is [Jaccard Distance|http://en.wikipedia.org/wiki/Jaccard_index],
> !http://upload.wikimedia.org/math/1/8/6/186c7f4e83da32e889d606140fae25a0.png!
> which we can estimate using technique called [MinHash|http://en.wikipedia.org/wiki/MinHash].
> In Cassandra, we can calculate and store MinHash signature when writing SSTable. New
compaction strategy uses the signature to find the group of similar SSTable for compaction
candidates. We can always fall back to STCS when such candidates are not exists.
> This is just an idea floating around my head, but before I forget, I dump it here. For
introduction to this technique, [Chapter 3 of 'Mining of Massive Datasets'|http://infolab.stanford.edu/~ullman/mmds/ch3.pdf]
is a good start.



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)

Mime
View raw message