Ramesh,
It is not easy to help you having only that much of information.
What I can assure you is that I personally had worked in the
development of a system where in production asserts more than 1 million
facts into working memory. So, it is not a soft/hard limitation in the
engine itself.
Your use case must be profiled to see where the memory is going.
Question: how much memory does your "ExtendedEvent" objects use? how
much the engine use? how much the rest of the application use?
On a rule writing tip, you realized by yourself that eval() is evil!!
and yet, you are not the only one that tried to use it... :) maybe we
should start a crusade against eval for people to be aware of how much
it hurts performance. :) We have to support it because some really
specific cases, need it, but it is not supposed to be used lightly.
So, going back to the point, 15000 facts is really not that much. But
the only way of understanding why you are having problems is profiling.
[]s
Edson
Reddy Ramesh-in2259c wrote:
Hi all,
This is Ramesh Reddy, trainee, in Motorola India Electronics pvt ltd.
I have a rule something like this....,
package com.motorola.nmatg.nms.nfm.suppression
import com.motorola.nmatg.nms.nfm.suppression.ExtendedEvent;
/*
This rule compress duplicates
*/
rule "Compress"
salience 10
when
event_1 : ExtendedEvent ( moid1 : source , severity1 : severity ,
eventType1 : eventType , eventTime1 : time )
event_2 : ExtendedEvent ( moid2 : source , severity2 : severity ,
eventType2 : eventType , eventTime2 : time )
eval (
moid1.equals(moid2) &&
severity1.intValue() == severity2.intValue() &&
eventType1.equals(eventType2) &&
eventTime2.longValue() >= eventTime1.longValue()
)
then
event_1.setStatus("compress");
event_2.setStatus("compress");
end
This is the rule which i am using in my DRL file. I am asserting
events into working memory such that all the events are matching with
the rule. But i am unable to assert more than around 1000 events into
workingMemory, I am keep asserting events into working memory, by the
time it reaches the count to around 1000 events, it is throwing memory
out of bound exception. I increased heap size from default 64mb to
512mb. with this i could able to assert more 300 events thats it.
when i got this problem, i modified my rule in the following way,
rule "Compress"
salience 10
when
event_1 : ExtendedEvent ( moid1 : source , severity1 : severity ,
eventType1 : eventType , eventTime1 : time )
event_2 : ExtendedEvent ( source == moid1 , severity == severity1 ,
eventType == eventType1 , time >= eventTime1 )
then
event_1.setStatus("compress");
event_2.setStatus("compress");
end
If i write my rule like this, then i could able to assert more 500
event objects. i.e i could able to assert in total around 1700 event
objects with this modified rule.
As per my understanding,
when we assert objects into working memory, pattern matching
is done at the time of assertion only, while this process, if we
create more variables in "when" part and using "eval" creating more
overhead(memory perspective) while pattern matching. When we assert
thousands of objects(events) into working memory, i think Drools
engine creates several objects internally(in rete network which i am
not aware), i think that would be the overhead.
But in my project, i need to work with around 15,000 event objects in
the working memory, of them, around 10,000 objects will match the rule.
Can any one in the group help me how to approach to solve this
problem. Is there anyway.. Ultimately, i have to work with 10000
events in working memory, given all the events matches with the rule
in DRL.
Thanks,
Ramesh
------------------------------------------------------------------------
*From:* [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] *On Behalf Of *Michael Neale
*Sent:* Tuesday, February 27, 2007 12:52 PM
*To:* Rules Users List
*Subject:* Re: [rules-users] How many objects we can assert into
working memorymaximum
The answer is:
42.
Or, perhaps, the length of a piece of string.
It really varies with how much memory you want to throw at it, thats
all really. The effectiveness of the RETE network may also add some
overhead.
I have heard of people pushing in millions of facts into one working
memory, but that would tend to be extreme - and it will use a lot of
memory.
IN 3.2 there are some added efficiencies over 3.0, when it comes out.
On 2/27/07, *Reddy Ramesh-in2259c* <[EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>> wrote:
Hi,
Is working memory size increases dynamically?? How many
objects we can assert into workingmemory? given, all the objects
are matching with rule in a drl file...
can anyone please tell me this...?
thanks,
Ramesh
_______________________________________________
rules-users mailing list
[email protected] <mailto:[email protected]>
https://lists.jboss.org/mailman/listinfo/rules-users
------------------------------------------------------------------------
_______________________________________________
rules-users mailing list
[email protected]
https://lists.jboss.org/mailman/listinfo/rules-users
--
Edson Tirelli
Software Engineer - JBoss Rules Core Developer
Office: +55 11 3124-6000
Mobile: +55 11 9218-4151
JBoss, a division of Red Hat @ www.jboss.com
_______________________________________________
rules-users mailing list
[email protected]
https://lists.jboss.org/mailman/listinfo/rules-users
