[infinispan-issues] [JBoss JIRA] (ISPN-3772) Reduce marshalling payload sizes
Galder Zamarreño (JIRA)
jira-events at lists.jboss.org
Thu Nov 28 11:39:06 EST 2013
[ https://issues.jboss.org/browse/ISPN-3772?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12927433#comment-12927433 ]
Galder Zamarreño commented on ISPN-3772:
----------------------------------------
I've played with arrays and I've noticed that trying to provide externalizer for primitive type arrays does produce smaller payloads. In fact, it produces payloads same size. However, for other non-primitive type arrays, it makes a difference. For example, given a list array (like the one used for segments) of containing two lists, each of one integer, would currently produce a payload of 40 bytes. With a List[] specific externalizer, that goes down to 22 bytes.
> Reduce marshalling payload sizes
> --------------------------------
>
> Key: ISPN-3772
> URL: https://issues.jboss.org/browse/ISPN-3772
> Project: Infinispan
> Issue Type: Enhancement
> Components: Marshalling
> Affects Versions: 6.0.0.Final
> Reporter: Galder Zamarreño
> Assignee: Galder Zamarreño
> Fix For: 6.1.0.Final
>
>
> Apply some marshalling improvements:
> 1. Marshall instances of java.util.Arrays$ArrayList with ArrayListExternalizer so that it occupies less space.
> 2. RegularEnumSet and JumboEnumSet should be detected and they should be written using SetExternalizer. For an enum set with a single value, this means going from writing 41 bytes to 10 bytes.
> 3. Avoid sending OPERATION_X flags when compatibility mode is disabled.
> 4. Current array support in JBoss Marshalling ends up writing class information even for know types. Try to figure out a way to detect the most commonly written arrays in Infinispan and find an efficient way to marshall them.
--
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
More information about the infinispan-issues
mailing list