[
https://issues.apache.org/jira/browse/HAMA-644?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Thomas Jungblut updated HAMA-644:
---------------------------------
Attachment: HAMA-644_1.patch
What I've done until now:
- reworked how the bundle works
- disk queue is now default, and backed by a linked list on disk instead of a
sequencefile (still experimental, we have to try out a few versions for polling
for performance comparision, currently solved by plain iteration)
- synchronized queues now have read/write locks
- some additional stuff
We are pretty verbose in the bundle now becaue we write the classname along
with each item. I hope this will be neglected by the compression, however I am
working on a linear grouping algorithm to bundle the messages with less memory
than grouping by a hashmap.
Take a closer look and tell me what you think. Be a little careful though,
because the bundle was one of our most used classes and the interface changed
as well as the internal behaviour.
> Messaging Scalability Improvement
> ----------------------------------
>
> Key: HAMA-644
> URL: https://issues.apache.org/jira/browse/HAMA-644
> Project: Hama
> Issue Type: Improvement
> Components: messaging
> Affects Versions: 0.5.0
> Reporter: Thomas Jungblut
> Fix For: 0.6.0
>
> Attachments: HAMA-644_1.patch
>
>
> more details follow soon
--
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
