[jboss-jira] [JBoss JIRA] (DROOLS-41) Activation Leakage using temporal windows

[Lorenzo Melegari (JIRA)]

     [ https://issues.jboss.org/browse/DROOLS-41?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Lorenzo Melegari updated DROOLS-41:
-----------------------------------

    Steps to Reproduce: 
You can find sample code in https://github.com/lome/lome_intext
Sample codes insert 1K Myfacts/seconds and aggregates them using the following rule:

rule "Dummy"
timer( cron:0/5 * * * * ? )
when
    Number( $count : intValue ) from accumulate( MyFact( ) over window:time(5s) from entry-point EventStream, sum(1) )
then
    System.out.println($count+" myfact(s) seen in the last 5 seconds");
end

After a while, the process goes out of memory.
[Profiler Screenshot|https://raw.github.com/lome/lome_intext/master/DroolsIssueWindow.png]
[Profiler Memory Inspect|https://raw.github.com/lome/lome_intext/master/DroolsIssueWindow_Mem.png]

As comparison, using the same event injector, this rule was also used:

declare MyFact
	@role( event )
	@timestamp( timestamp )
	@expires( 5s )
end

rule "Dummy"
timer( cron:0/5 * * * * ? )
when
	Number( $count : intValue ) from accumulate( MyFact( ) from entry-point EventStream, sum(1) )
then
    System.out.println($count+" myfact(s) seen in the last 5 seconds");
end

Reaching the same result of aggregating over a fixed time window, but, this time, the memory consumption and usage turns out to be way better

[Profiler Screenshot|https://raw.github.com/lome/lome_intext/master/DroolsIssue_NoWindow.png]
[Profiler Memory Inspect|https://raw.github.com/lome/lome_intext/master/DroolsIssueNoWindow_Mem.png]

  was:
You can find sample code in https://github.com/lome/lome_intext
Sample codes insert 1K Myfacts/seconds and aggregates them using the following rule:

rule "Dummy"
timer( cron:0/5 * * * * ? )
when
    Number( $count : intValue ) from accumulate( MyFact( ) over window:time(5s) from entry-point EventStream, sum(1) )
then
    System.out.println($count+" myfact(s) seen in the last 5 seconds");
end

After a while, the process goes out of memory.
[Profiler Screenshot|https://raw.github.com/lome/lome_intext/master/DroolsIssueWindow.png]
[Profiler Memory Inspect|https://raw.github.com/lome/lome_intext/master/DroolsIssueWindow.png]

As comparison, using the same event injector, this rule was also used:

declare MyFact
	@role( event )
	@timestamp( timestamp )
	@expires( 5s )
end

rule "Dummy"
timer( cron:0/5 * * * * ? )
when
	Number( $count : intValue ) from accumulate( MyFact( ) from entry-point EventStream, sum(1) )
then
    System.out.println($count+" myfact(s) seen in the last 5 seconds");
end

Reaching the same result of aggregating over a fixed time window, but, this time, the memory consumption and usage turns out to be way better

[Profiler Screenshot|https://raw.github.com/lome/lome_intext/master/DroolsIssue_NoWindow.png]
[Profiler Memory Inspect|https://raw.github.com/lome/lome_intext/master/DroolsIssueNoWindow_Mem.png]


    
> Activation Leakage using temporal windows
> -----------------------------------------
>
>                 Key: DROOLS-41
>                 URL: https://issues.jboss.org/browse/DROOLS-41
>             Project: Drools
>          Issue Type: Bug
>      Security Level: Public(Everyone can see) 
>    Affects Versions: 5.5
>         Environment: Sun JDK 1.6
>            Reporter: Lorenzo Melegari
>            Assignee: Mark Proctor
>              Labels: memory_leak
>
> When aggregating over time windows on frequent events (about 1K events per second), lot of EventFactHandle are generated and retained in memory. 

--
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