4:36 p.m.
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 ?
7:16 p.m.
New subject: Drools 4 performs slower then Drools 3 - Help on .drlfile adaptation
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);
From: rules-users-bounces(a)lists.jboss.org [mailto:rules-users-bounces@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 ?
2:56 p.m.
New subject: Drools 4 performs slower then Drools 3 - Help on .drlfile adaptation
Hi Joe,
On Wed, Nov 19, 2008 at 8:16 PM, Joe White <Joe.White(a)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@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(a)lists.jboss.org [mailto:
rules-users-bounces(a)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(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users
--
Cheers,
Maxim Veksler
"Free as in Freedom" - Do u GNU ?
12:03 a.m.
New subject: Drools 4 performs slower then Drools 3 - Help on .drlfile adaptation
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(a)gmail.com>
Hi Joe,
On Wed, Nov 19, 2008 at 8:16 PM, Joe White <Joe.White(a)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@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(a)lists.jboss.org [mailto:
> rules-users-bounces(a)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(a)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(a)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
5641
days inactive
5643
days old
3 comments
3 participants
participants (3)
-
Edson Tirelli -
Joe White -
Maxim Veksler