cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Alex Petrov (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-12390) Make SASI work with partitioners that have variable-size tokens
Date Fri, 18 Aug 2017 07:27:00 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-12390?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16131847#comment-16131847
] 

Alex Petrov commented on CASSANDRA-12390:
-----------------------------------------

Just to give the heads up, in my opinion this makes sense to have when [CASSANDRA-11990] lands,
as it already does quite some binary format changes.

The main problem with variable token sizes is the fact that none of the current offset mechanisms
would work and would most mean a major slowdown (we did a prototype back in the day), so having
row offsets can win some of that performance back.

If we do it strictly for the byte ordered partitioner support, I'd say unless there's a good
way to avoid performance hit on seeks, might be better to avoid it.


> Make SASI work with partitioners that have variable-size tokens
> ---------------------------------------------------------------
>
>                 Key: CASSANDRA-12390
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-12390
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: sasi
>            Reporter: Alex Petrov
>            Assignee: Abhish Agarwal
>
> At the moment, SASI indexed can work only with Murmu3Partitioner. [CASSANDRA-12389] was
created to enable support of one more partitioner with fixed-size tokens, although enabling
variable-size tokens will need more work, namely skipping tokens, since we can't rely on fixed-size
multiplication for calculating offsets in that case anymore.
> This change won't require bytecode format changes, although supporting ByteOrderedPartitioner
is not a very high priority, and performance will be worse because of "manual" skipping. 



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@cassandra.apache.org
For additional commands, e-mail: commits-help@cassandra.apache.org


Mime
View raw message