[ 
https://issues.apache.org/jira/browse/CASSANDRA-4858?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13552588#comment-13552588
 ] 

Sylvain Lebresne commented on CASSANDRA-4858:
---------------------------------------------

bq. but scores are really helped in the past returning the queries within SLA's

I agree, and the goal of the patch is certainly not to deny that. The question 
is, if you can merge 2 queries into 1, at which point do you decide it's not 
worth it based on the endpoints scores. I think it is safe to say that in most 
case, picking 1 query over 2 will be the best option. That being said, if you 
do happen to have a very very slow node that the intersection would pick while 
doing 2 query would only pick very fast nodes, then you might be worth off 
doing the 1 query. So along with a rebase of the previous patch I'm attaching a 
2nd patch that adds an heuristic to the dynamic snitch that tries to avoid 
that. The heuristic being that if the average score of the intersection of 2 
replica set is twice as bad (I'll admit that the 'twice' here is a bit random 
and might not be appropriate) as the average score of one of the replica set, 
then we ignore the intersection. It's probably not the best heuristic though, 
but it has the advantage of being simple.
                
> Coverage analysis for low-CL queries
> ------------------------------------
>
>                 Key: CASSANDRA-4858
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-4858
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: Core
>            Reporter: Jonathan Ellis
>            Assignee: Vijay
>             Fix For: 1.2.1
>
>         Attachments: 0001-CASSANDRA-4858.patch, 0001-CASSANDRA-4858-v2.patch, 
> 4858-v3-1.txt, 4858-v3-2.txt
>
>
> There are many cases where getRangeSlice creates more
> RangeSliceCommand than it should, because it always creates one for each range
> returned by getRestrictedRange.  Especially for CL.ONE this does not take
> the replication factor into account and is potentially pretty wasteful.
> A range slice at CL.ONE on a 3 node cluster with RF=3 should only
> ever create one RangeSliceCommand.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Reply via email to