[ 
https://issues.apache.org/jira/browse/SAMOA-51?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15052740#comment-15052740
 ] 

ASF GitHub Bot commented on SAMOA-51:
-------------------------------------

Github user senorcarbone commented on the pull request:

    https://github.com/apache/incubator-samoa/pull/44#issuecomment-163935565
  
    I tested VHT with 8 local workers and rather rather sufficient allocated 
memory (4GB) and it still takes a while, around 1m50sec. There will be an 
additional patch soon, probably collocated with Flink v1.0 that will address 
these performance issues on iterative streaming jobs. 
    
    <img width="1433" alt="screen shot 2015-12-11 at 14 21 17" 
src="https://cloud.githubusercontent.com/assets/858078/11744710/e16561e0-a012-11e5-88dc-843f2c1a55ed.png";>



> Update Flink Module to v0.10.0
> ------------------------------
>
>                 Key: SAMOA-51
>                 URL: https://issues.apache.org/jira/browse/SAMOA-51
>             Project: SAMOA
>          Issue Type: Task
>          Components: SAMOA-Flink
>            Reporter: Paris Carbone
>            Priority: Minor
>
> Flink v0.10.0 is fresh and comes with many performance and api improvements 
> so an upgrade is a must! 
> The only non-trivial thing that needs change is, again, loops. Flink now  
> assumes a forward parallelism perseverance between components. In other 
> words, we need to repartition sources to match the tail parallelism before 
> starting a flink iteration.
> I will hopefully have a fully working and tested version this week 



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to