[
https://issues.apache.org/jira/browse/CASSANDRA-16717?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17567948#comment-17567948
]
Alex Petrov commented on CASSANDRA-16717:
-----------------------------------------
Unless we add a significant amount of fuzz tests for column filter, I would not
be very keen to include new optimisations, since some of them have led to
significant issues in the past. Allowing for a difference between different
consistency levels may make it even more likely to introduce a hard-to-see
issues.
> The fetching strategy can be optimized for CL.ONE and CL.LOCAL_ONE
> ------------------------------------------------------------------
>
> Key: CASSANDRA-16717
> URL: https://issues.apache.org/jira/browse/CASSANDRA-16717
> Project: Cassandra
> Issue Type: Improvement
> Reporter: Benjamin Lerer
> Assignee: Benjamin Lerer
> Priority: Normal
>
> The current {{ColumnFilter}} fetching strategy has been implemented in order
> to guaranty the CQL semantics around empty vs non-existing rows. It has also
> some importance regarding read-repair (CASSANDRA-16710). Nevertheless, reads
> at {{CL.ONE}} and at {{CL.LOCAL_ONE}} do not use read-repair and do not
> require more columns than the queried ones as those cannot be deleted by
> column deletions present on other nodes. By consequence, having
> {{ColumnFilter}} fetching only the queried columns should improve query
> speed, by reducing the number of SSTable reads for queries fetching specific
> rows and reducing the amound of data serialized and transfered between nodes
> (if the data is not local).
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
