[rules-users] Pooling stateful sessions to have threaded executions of a ruleset

[Maxime Falaize]

I'm sorry but I don't understand what you are talking about. In this
article, there is nothing about multithreaded rules evaluation. It just
says that the Phreak algorithm is designed for thread safety and
*future*multicore processors exploitation. It doesn't mean that it is
already
multithreaded. Moreover I tested my appliciation with and without
multithreaded sessions (I am using Drools 6.0.1.Final) and I noted a faster
execution in the multithreaded one.


2014-04-04 13:53 GMT+02:00 jmercier <jmercier at genoscope.cns.fr>:

> Le 04/03/2014 09:55, Maxime Falaize a écrit :
> > Hello,
> >
> > I want to ask you if it is a good practive to pool stateful sessions
> > for a specific ruleset to improve the execution performance.
> > Actually in my application I execute my rules by calling SOAP
> > webservice. For performance purpose, I test multithreaded calls to my
> > webservice and I noted that when I pool sessions in the server side,
> > it improves the performance a lot.
> >
> > To pool sessions, I just declare multiple ksession tag in my
> > kmodule.xml :
> >
> > <kbase name="KBase" packages="com.example.*">
> >         <ksession name="KSession1"/>
> >          <ksession name="KSession2"/>
> >         <ksession name="KSession3"/>
> >         <ksession name="KSession4"/>
> >         <ksession name="KSession5"/>
> >  </kbase>
> >
> > In my spring webservice endpoint I just put that code to handle the
> > pool :
> >
> > @Endpoint
> > public class ExampleEndpoint implements InitializingBean {
> >
> >     @Autowired
> >     private ExampleRuleService ruleService;
> >      private Map<Integer, Boolean> isRunningMap = new
> > HashMap<Integer, Boolean>();
> >     private static final int NB_POOL_SESSIONS = 5;
> >
> >     @PayloadRoot(localPart = "com.example.ExampleRequest")
> >      @ResponsePayload
> >     public ExampleResponse handleRequest(
> >             @RequestPayload ExampleRequest request) throws
> > InterruptedException {
> >         KieServices ks = KieServices.Factory.get();
> >         KieContainer kc = ks.getKieClasspathContainer();
> >          while (true) {
> >             for (int i = 0; i < NB_POOL_SESSIONS; i++) {
> >                 boolean run = false;
> >
> >                 synchronized (isRunningMap) {
> >                     if (!isRunningMap.get(i)) {
> >                          isRunningMap.put(i, true);
> >                         run = true;
> >                     }
> >                 }
> >
> >                 if (run) {
> >                     KieSession ksession = kc.newKieSession("KSession"
> > + (i + 1));
> >                      ExampleResponse response =
> > ruleService.run(ksession, request);
> >                     ksession.dispose();
> >
> >                     isRunningMap.put(i, false);
> >                     return response;
> >                  }
> >             }
> >             Thread.sleep(100);
> >         }
> >     }
> >
> >     public void afterPropertiesSet() throws Exception {
> >         for (int i = 1; i <= NB_POOL_SESSIONS; i++) {
> >             isRunningMap.put((i - 1), false);
> >          }
> >     }
> >
> > }
> >
> > It works well because in my benchmark I improve 5 times the
> > performance (as I have 5 different threads) but I wondered if it is a
> > good practice and if it does not hide any issues that I could have in
> > the future.
> >
> > Thanks for your help.
> >
> > --
> > Maxime FALAIZE
> > _______________________________________________
> > rules-users mailing list
> > rules-users at lists.jboss.org
> > https://lists.jboss.org/mailman/listinfo/rules-users
>
>
> Hi maxime,
>
> I do not remember if use drools 6 or drools 5. If you using drools 6.
> Phreaks algorithm use multi threading according to use 'from
> accummulate' far i understand here:
> http://planet.jboss.org/post/drools_6_performance_with_the_phreak_algorithm
>
> Instead to put a thread by ksession here rules evaluation are
> multi-threaded.
>
>
> _______________________________________________
> rules-users mailing list
> rules-users at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/rules-users




-- 
Maxime FALAIZE
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.jboss.org/pipermail/rules-users/attachments/20140407/83c8f278/attachment.html