Drools performance as a rule engine using multiple threads

Question

I'm using a single threaded drools project as a first-match policy/condition evaluation engine. I'm using the following code to get a kieSession from the kieContainer, and evaluate the predicate based on the conditions.

KieContainer currentContainer = kieContainer.get();
StatelessKieSession newKIESession = currentContainer.newStatelessKieSession();
newKieSession.execute(predicate);

The evaluation takes about 10 ms for a given request. However, to increase the performance when I used 10 threads, each of these requests starting taking about 100ms effectively giving me the same performance with a larger number of threads.

1. What is the RIGHT WAY to use multi-threading to improve condition evaluation in Drools?

2. Should I spawn a new container for each processing thread instead spawning of a new session from existing containers?

   //SEE LAST LINE BELOW
   
   KieServices ks = KieServices.Factory.get();
   KieRepository kr = ks.getRepository();
   KieFileSystem kfs = ks.newKieFileSystem();
   
   byte[] drlAsBytes = retrieveDRLResource();
   kfs.write(ResourceFactory.newByteArrayResource(drlAsBytes).setTargetPath(DROOLS_DEFAULT_PATH));
   
   KieBuilder kb = ks.newKieBuilder(kfs);
   
   kb.buildAll(); // kieModule is automatically deployed to KieRepository if successfully built.
   
   if (kb.getResults().hasMessages(Level.ERROR)) {
       throw new PolicyServiceException("Build Errors:\n" + kb.getResults().toString());
   }
   
   //******* SHOULD THIS BE DONE FOR EACH PROCESSING THREAD? ******
   KieContainer newkieContainer = ks.newKieContainer(kr.getDefaultReleaseId());
   

I'm using the following version of DROOLS.

    <dependency>
        <groupId>org.drools</groupId>
        <artifactId>drools-compiler</artifactId>
        <version>7.2.0.Final</version>
    </dependency>