[infinispan-issues] [JBoss JIRA] Commented: (ISPN-936) putAll suboptimal and possibly wrong

[Manik Surtani (JIRA)]

    [ https://issues.jboss.org/browse/ISPN-936?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12585889#comment-12585889 ] 

Manik Surtani commented on ISPN-936:
------------------------------------

Actually this is a questionable optimisation.  putAll() with a large map of small entries can be wasteful, but then a small map - say 3 large entries - may be better off sending this around everywhere since it would be very costly to serialize the map 3 times over for 3 different recipients, each with slightly different contents.

At the end of the day, the putAll() API is purely to satisfy java.util.Map.  If you are using large complex objects or are performance sensitive, use put(K, V) instead. :-)

> putAll suboptimal and possibly wrong
> ------------------------------------
>
>                 Key: ISPN-936
>                 URL: https://issues.jboss.org/browse/ISPN-936
>             Project: Infinispan
>          Issue Type: Feature Request
>          Components: Distributed Cache
>    Affects Versions: 4.2.0.Final
>            Reporter: Mircea Markus
>            Assignee: Manik Surtani
>
> DistributionManager.visitPutMapCommand broadcasts ALL the map entries to ALL the nodes to which keys are mapped.
> E.g. if Map={k1=v1, k2=v2} and k1 maps to node A, k2 maps to node B then both A and B would receive {k1=v1, k2=v2}. 
> The way it should be is A receives {k1=v1} and B receives {k2=v2}. 
> This is suboptimal and error prone as, by the look of code, both nodes would store ALL the data in the data container. 
>   

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira