[rules-users] Drools 4 performs slower then Drools 3 - Help on .drlfile adaptation

[Edson Tirelli]

   Maxim,

   There is a lot to look for when talking about performance. It is not
possible to give you any advice without knowing explicitly what you are
doing. It's like SQL: depending one your data model and the way you write
your queries, they will perform better or worst.

   Just one detail: most of the times, this will reduce your performance:

conf.setIndexRightBetaMemory(false);

   Obviously, we expose the configuration because in a few very specific
cases, it may improve it.

   So, if you really want to discuss how to improve performance, create a
self contained test case and present it. Then we have the data to talk about
it.

   Edson


2008/11/20 Maxim Veksler <maxim.veksler at gmail.com>

> Hi Joe,
>
> On Wed, Nov 19, 2008 at 8:16 PM, Joe White <Joe.White at recondotech.com>wrote:
>
>>  Maxim,
>>
>> Depending on your use case you may be able to disable shadow facts in your
>> Drools 4 instance. See the manual on when it is ok/not-ok to disable shadow
>> facts. Based on your description it probably is acceptable for your uses.
>> Disabling shadow facts may improve your performance, the manual has details.
>>
>>
>>
>> To disable shadow facts:
>>
>>   RuleBaseConfiguration conf = *new* RuleBaseConfiguration();
>>
>>                     conf.setShadowProxy(false);
>>
>>                     base = RuleBaseFactory.*newRuleBase*(conf);
>>
>>
>>
>
> Thank you for the tip, I've added the following flags :
>
>         conf.setShadowProxy(false);
>         conf.setAlphaMemory(false);
>         conf.setIndexLeftBetaMemory(false);
>         conf.setIndexRightBetaMemory(false);
>
> and yet I'm seeing performance that are not near what I've been expecting,
> am I doing something wrong or does drools gives me all that he can :
>
> [maximv at mldev2 ~]$ awk -f awk_avg_ruleEngineFacade.getMatch.awk
> /var/log/Manager.log
> avg: 35.9153(ms) for 1027 calls
>
>
> This is an example of the kind of rules we're using (modified) :
>
> rule "Context OtherObject"
> when
>     $f : FilteringStateConnector($ContextId : ContextId, $subscriberId :
> subscriberId)
>
>     $subscriber : Subscriber(id == $subscriberId, $subscriberOtherObject :
> OtherObject, OtherObject != null , OtherObject != "")
>     $OtherObject : PropertyInstance(type_ ==
> PropertyInstance.OtherObject_PROPERTY_TYPE, code ==
> $subscriberOtherObject)
>     $Context : Context(code == $ContextId, OtherObjects contains
> $OtherObject)
>
> then
>     $f.setPassedRuleRecord("Context OtherObject");
> end
>
>
>
>>  *From:* rules-users-bounces at lists.jboss.org [mailto:
>> rules-users-bounces at lists.jboss.org] *On Behalf Of *Maxim Veksler
>> *Sent:* Wednesday, November 19, 2008 8:36 AM
>> *To:* Rules Users List
>> *Subject:* [rules-users] Drools 4 performs slower then Drools 3 - Help on
>> .drlfile adaptation
>>
>>
>>
>> Hello everyone,
>>
>>
>> Please allow me to briefly describe how our system works:
>>
>> We use drools as an efficient matching layers - I have X rules, in the RHS
>> of each rule I have a mark for each rule to signal that it has passed in the
>> current context (There are by definition more then 1 contexts). After the
>> first fireAllRules() has finished, and I have the data structure that allows
>> me to query what rules have passed for each context I assert a "Marker"
>> object, that allows Drools to continue filtering only for these contexts
>> that I decide as valid to match attempts.
>>
>> The end result, is that I have 2 fireAllRules called with some java code
>> in between them.
>>
>>
>> In drools3 for 3 contexts I got the following resutls :
>>
>> == Drools Section (Generic Call) : With default context settings, no
>> HotSpot.
>> RESULT::::::::    avg: 33.9023(ms) for 1555 calls
>>
>> Now in drools4 I'm getting the following:
>>
>> == Drools Section (Generic Call) : With default context settings, no
>> HotSpot - Drools 4
>> RESULT::::::::    avg: 37.4356(ms) for 1467 calls
>>
>>
>>
>> As I increase the context's number I'm getting larger difference in
>> results time.
>> Please note that HotSpot is disabled just for the test, to not JIT
>> interference.
>>
>> I've disabled all evals in the drl files, to all is left is pure Drools
>> syntax and yet It seems that the system works slower.
>>
>> Please suggest what else I should be doing to optimize the drl files to
>> work efficiently with Drools 4 engine.
>>
>>
>> Thank you,
>> Maxim.
>>
>> --
>> Cheers,
>> Maxim Veksler
>>
>> "Free as in Freedom" - Do u GNU ?
>>
>> _______________________________________________
>> rules-users mailing list
>> rules-users at lists.jboss.org
>> https://lists.jboss.org/mailman/listinfo/rules-users
>>
>>
>
>
> --
> Cheers,
> Maxim Veksler
>
> "Free as in Freedom" - Do u GNU ?
>
> _______________________________________________
> rules-users mailing list
> rules-users at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/rules-users
>
>


-- 
 Edson Tirelli
 JBoss Drools Core Development
 JBoss, a division of Red Hat @ www.jboss.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.jboss.org/pipermail/rules-users/attachments/20081121/21c4462a/attachment.html