Drools, jBPM OptaPlanner presentation

1. • Distinguished Engineer II
• Chief Architect
• Projects : Drools, jBPM and OptaPlanner
• Release per sprint
• 3 weeks per sprint
• Products : Red Hat PAM (BPMS) / DM (BRMS)
• Release per quarter
Mark Proctor

2. The Journey
2

3. • Drools (early 2000)
The Journey
<import>org.drools.examples.fibonacci.Fibonacci</import>
<rule name="Bootstrap 1" salience="20">
<parameter identifier="f">
<class>Fibonacci</class>
</parameter>
<java:condition>f.getSequence() == 1</java:condition>
<java:condition>f.getValue() == -1</java:condition>
<java:consequence>
f.setValue( 1 );
System.err.println( f.getSequence() + " == " + f.getValue() );
drools.modifyObject( f );
</java:consequence>
</rule>
Drools 2.x XML Syntax

4. The Journey
• Drools (early 2000)
• Drools 3 (2006)
• Joined JBoss 2005
• RHT acquired JBoss in 2006
import org.drools.examples.fibonacci.FibonacciExample.Fibonacci;
rule Bootstrap when
f : Fibonacci( sequence == 1 )
then
modify ( f ){ value = 1 };
System.out.println( f.sequence + " == " + f.value );
end
Drools 3.x - 7.x Syntax

5. The Journey
• Drools (early 2000)
• Drools 3 (2006)
• Drools 4 (2007)
• Increased expressiveness
• Better Performance
• Basic Drools Flow
• Basic BRMS

6. The Journey
• Drools (early 2000)
• Drools 3 (2006)
• Drools 4 (2007)
• Drools 5 (2009)
• BRMS improvements
• CEP
• Drools-Flow full workflow with BPMN

7. • Drools (early 2000)
• Drools 3 (2006)
• Drools 4 (2007)
• Drools 5 (2009)
• jBPM 5 (2011)
• incorporated from drools-flow
The Journey

8. The Journey
• Drools (early 2000)
• Drools 3 (2006)
• Drools 4 (2007)
• Drools 5 (2009)
• jBPM 5 (2011)
• Polymita acquisition (2012)

9. The Journey
• Drools (early 2000)
• Drools 3 (2006)
• Drools 4 (2007)
• Drools 5 (2009)
• jBPM 5 (2011)
• Polymita acquisition (2012)
• Drools & jBPM 6.0 (2013)
• Cloud runnable
• Modular Workbench
• Lazy rule algorithm
• Methodology based:
• Author, build, deploy, utilise
• Declarative
• OptaPlanner

10. The Journey
• Drools (early 2000)
• Drools 3 (2006)
• Drools 4 (2007)
• Drools 5 (2009)
• jBPM 5 (2011)
• Polymita acquisition (2012)
• Drools & jBPM 6.0 (2013)
• Drools & jBPM 7.0 (2018)
• Cloud enabled
• Improved UXD around a core set of functionality
• Targeting the Business Practitioner.
• Case Management

11. The Journey
• Drools (early 2000)
• Drools 3 (2006)
• Drools 4 (2007)
• Drools 5 (2009)
• jBPM 5 (2011)
• Polymita acquisition (2012)
• Drools & jBPM 6.0 (2013)
• Drools & jBPM 7.0 (2018)
• Submarine (2019)

12. The Journey
• Drools (early 2000)
• Drools 3 (2006)
• Drools 4 (2007)
• Drools 5 (2009)
• jBPM 5 (2011)
• Polymita acquisition (2012)
• Drools & jBPM 6.0 (2013)
• Drools & jBPM 7.0 (2018)
• Submarine (2019)
• Since 2011 quadrupled engineering
team size (2019)
• with dedicated QE (2.5 QE to 1 Eng)
and UXD.

13. • Drools (early 2000)
• Drools 3 (2006)
• Drools 4 (2007)
• Drools 5 (2009)
• jBPM 5 (2011)
• Polymita acquisition (2012)
• Drools & jBPM 6.0 (2013)
• Drools & jBPM 7.0 (2018)
• Submarine (2019)
• Since 2011 quadrupled engineering team size (2019)
• Super Stars
The Journey

14. And Then This Happened
14

17. RHT Cloud
17

18. ● Multi Cloud
● Using multiple public clouds.
● Avoid vendor lock-in.
● Lower costs, via competition
Multi Cloud

19. ● Hybrid Cloud.
● Combination of 1 or more public and private clouds.
● With a degree of portability.
Hybrid Cloud

20. ● Virtual Machines (Early Cloud).
● Virtualises the hardware.
● Slow startup.
● Inefficient use of resources (uses all memory).
● Containerisation (Modern Cloud) aka Kubernetes.
● Virtualises the Operating System.
● Instead of providing virtual hardware to a VM, you provide a virtual OS to your
application.
● Run large number of applications, with fast startup and efficient resource
utilisation.
● Higher Density.
Virtual Machines vs Containerisation

21. ● Kubernetes is the project.
● OpenShift is the RHT Product based on Kubernetes.
● OpenShift has:
● Stricter security policies.
● Integrated CI/CD with Jenkins.
● Image Streams for image management.
● Single image with managed virtual tags (dev/stage/prod)
● Automated redeployment on new versions
● Automate build chains
OpenShift vs Kubernets

22. KNative
Kubernetes-based platform to build, deploy, and manage modern serverless workloads.
● KNative Build
○ Configurable and flexible approach to building source code into containers
● KNative Serving
○ Rapid deployment of serverless containers, with automatic scale-to-zero support.
○ Revisions mark point-in-time snapshots of deployed code and configurations.
● KNative Eventing
○ Universal subscription, delivery, and management of events
○ Scalable from just a few events to live streams

23. ● Stateless (with Ephemeral Storage)
● Push caching and scaling to OCP K8
● Fast creation of pods on demand
● Push state external
● Sticky Sessions
● Templates and Operators
● Istio (Service Mesh)
● KNative / Serverless
Cloud Native Application

24. Quarkus (GraalVM)
24

25. ● A polyglot VM with cross-language JIT supporting:
○ Java Bytecode and JVM languages
○ Interop with different languages
○ Dynamic languages through Truffle API
● Cross-language interop out of the box
○ Simple AST-based interpreter
○ JIT across language boundaries
● Support for native binary compilation (SubstrateVM)
○ faster boot-up
○ lower memory footprint

26. AOT Compilation
● Static Analysis
● Closed World Assumption
● Dead Code elimination
Faster Process
Start Time
Lower Memory
Footprint
Smaller Size
On Disk

27. ● A Kubernetes Native Java stack tailored for GraalVM & OpenJDK
HotSpot
Eclipse Vert.x Hibernate RESTEasy Apache Camel Eclipse MicroProfile Netty
Kubernetes OpenShift Jaeger Prometheus Apache Kafka Infinispan

28. A cohesive platform for optimized developer joy:
! Based on standards, but not limited
! Unified configuration
! Zero config, live reload in the blink of an eye
! Streamlined code for the 80% common usages,
flexible for the 20%
! No hassle native executable generation
Developer Joy

29. Memory (RSS) in Megabytes
Quarkus + GraalVM
13 MB
Quarkus + OpenJDK
74 MB
Traditional Cloud Native Stack
140 MB
REST
Supersonic Subatomic Java

30. Memory (RSS) in Megabytes
REST +
CRUD
Quarkus + GraalVM
35 MB
Quarkus + OpenJDK
130 MB
Traditional Cloud Native Stack
218 MB
Supersonic Subatomic Java

31. Boot + First Response Time (in seconds)
Quarkus + GraalVM 0.014 Seconds
REST
REST +
CRUD
Quarkus + OpenJDK 0.75 Seconds
Quarkus + GraalVM 0.055 Seconds
Quarkus + OpenJDK 2.5 Seconds
Traditional Cloud Native Stack 9.5 Seconds
Traditional Cloud Native Stack 4.3 Seconds
Time to first response
Supersonic Subatomic Java

32. Submarine
32

33. ○ Community, Skunk Works, effort exploring a number of directions
○ Cloud native (Kubernetes) architecture
■ First class environment for development of applications on OCP.
○ GraalVM / Quarkus
○ Domain-specific REST process service
○ OmniChannel tooling with VSCode/Che
■ With consistent Business Practitioner version (e.g. electron.js)
○ Goal - provide services* attractive to OpenShift application developers for:
■ Decisions (rules)
■ Process (workflow)
■ Events
■ Optimizations
Submarine

34. ● A pure Java DSL for rules and process authoring
● A pure Java canonical representation of a rule base
● Automatically generated by Maven plugin or Quarkus extension
● Can be embedded in jar
● Faster boot
● Allow for faster prototyping and experimentation of new features, without
worrying about language design
● Prerequisite to make Drools and jBPM natively compilable on GraalVM
Canonical Executable Model : What

35. ● Do not parse BPMN, DRL, DMN etc at runtime.
○ Generate code from parsers at build time.
○ Executable means java rules and workflow fluent DSLs.
● Removed
○ Reflection
○ Bytecode Generation
● Move all “determinism” to the language layer
○ Indexes, property reactivity
Canonical Executable Model : How

36. Canonical Executable Model : Rules

37. Canonical Executable Model : Processes

38. Canonical Executable Model : drools.js
const Person = Java.type('org.submarine.examples.polyglot.Person');
const markV = declarationOf( Person.class );
const olderV = declarationOf( Person.class );
const r = Rule("X is older than Mark").build(
Pattern(markV).expr(p => p.getName() === "Mark"),
Pattern(olderV)
.expr(p => p.getName() !== "Mark")
.expr(markV, (p1, p2) => p1.getAge() > p2.getAge()),
on(olderV, markV).execute((p1, p2) =>
console.log( p1.getName() + " is older than " + p2.getName())));
const m = Model(r);
const s = Session(m);
console.log("start")
s.insert(new Person("Mark", 37));
s.insert(new Person("Edson", 35));
s.insert(new Person("Mario", 40));
s.fireAllRules();
console.log("end")

39. Canonical Executable Model : jbpm.js
const ProcessRuntime = Java.type("org.kie.api.process.ProcessRuntime")
const RuleFlowProcessFactory = Java.type('org.jbpm.ruleflow.core.RuleFlowProcessFactory');
const factory = RuleFlowProcessFactory.createProcess("org.kie.api2.MyProcessUnit");
factory
// Header
.name("HelloWorldProcess").version("1.0").packageName("org.jbpm")
// Nodes
.startNode(1).name("Start").done()
.actionNode(2).name("Action")
.action(ctx => {
console.log("this is an actionprotean");
}).done()
.endNode(3).name("End").done()
// Connections
.connection(1, 2)
.connection(2, 3);
p = factory.validate().getProcess();

40. Onboarding
Service
Payroll
Service
HR
Service
https://github.com/mswiderski/onboarding-sample
Domain-specific Service: Onboarding application

43. OmniChannel Architecture

44. Architecture

47. Drools Challenges
47

48. OPS
48
(p mb17
(goal ^status active ^type on ^object <o>)
(object ^name <o> ^at <p>)
(monkey ^at <p> ^holds nil)
-->
(write (crlf) "climb onto" <o>)
(modify 3 ^on <o>)
(modify 1 ^status satisfied))
OPS
•Official Production System.
•pre OPS5 linear Search.
•Conflict Resolution
McDermott, Charles L (Charles Lanny) Forgy.
“Production System Conflict
Resolution Strategies,” 1976, 23.
OPS5 1981
•Language
•Rete
•LEX/MEA Working
Memory
Rules
Inference Engine
Pattern
Matching
Agenda

49. Jess
49
(deftemplate employee (slot salary) (slot name))
(defrule count-highly-paid-employees
?c <- (accumulate (bind ?count 0) ;; initializer
(bind ?count (+ ?count 1)) ;; action
?count ;; result
(employee (salary ?s&:(> ?s 100000)))) ;; CE
=>
(printout t ?c " employees make more than $100000/year." crlf))
Jess
•Java implementation of Clips syntax
Jess7 2006
•Accumulate
•Slot Specific
•Backward Chaining (ish)
Working
Memory
Modules
Inference Engine
Pattern
Matching
Agenda

50. Drools
50
Drools
•Clips/Jess derivative with Java-like language
Drools 7.x
•Accumulate
•Temporal Operators
•Property Reactive
•Prolog(ish) Backward Chaining
•OOPath
Working
Memory
Modules
Inference Engine
Pattern
Matching
Agenda

51. Still Here
51
Working MemoryModules
Inference Engine
Pattern Matching
Agenda

52. Still Here
52
Working
MemoryModules
Inference Engine
Pattern
Matching
Agenda
“Today, I want to make clear that the future prospects for production
rule technology are diminishing..…
production rule technology is inadequate ”
(Paul Haley 2013)
http://haleyai.com/wordpress/2013/06/22/confessions-of-a-production-rule-vendor-part-1/

53. Still Here
53
Working
MemoryModules
Inference Engine
Pattern
Matching
Agenda
“Today, I want to make clear that the future prospects for production
rule technology are diminishing..…
production rule technology is inadequate”
(Paul Haley 2013)
http://haleyai.com/wordpress/2013/06/22/confessions-of-a-production-rule-vendor-part-1/

54. Active DataBases
•Validation enforcement, triggers.
•Simpler semantics, no forward chaining.
•Without 2 phase. Does not allow mutation of selected tables.
•Scoped deterministic execution within the transaction
Decisioning
•DMN
CEP
•Power of production system LHS, with additional temporal reasoning
•No inference, rules are isolated
Reactive Programming with Pattern Matching
•Light Weight, simple facet of a production system
•No Reactive Joins, but can do passive joins
•Built into native language
Offshoots
54

55. Reactive Programming / Pattern Matching
55
public static class Customer {
int String name;
int discount = 0;
}
public static class GoldCustomer extends Customer { }
public static class SilverCustomer extends Customer { }
public static class BronzeCustomer extends Customer { }
public static void main( String[] args ) {
observable.subscribe( customer -> {
switch (customer) {
case GoldCustomer g: {
CriteriaQuery<Double> cr = cb.createQuery(Double.class);
Root<ShoppingCart> root = cr.from(ShoppingCart.class);
cr.select(cb.sum(root.get("items.value"))).where("customer = " + g.name);
Query<Double> query = session.createQuery(cr);
double total = query.getResultList().get(0);
g.discount = total > 100 ? 20 : 15;
break;
}
case SilverCustomer s: s.discount = 10; ………… break;
case BronzeCustomer b: b.discount = 5; ………… break;
default: customer.discount = 0;
}
} );
}

56. Blended Languages - LinQ / Reactive LinQ
56
var result=
(
from data in myData
join inData in incomingData
on data.ID equals inData.ID
select new
{
data.ID,
inData.LOCATION,
data.COUNT
}
).ToList();

57. Still Here
57
Working
MemoryModules
Inference Engine
Pattern
Matching
Agenda
“It’s time to trade rule technology dating back to the 80’s for state of
the art AI…
Artificial intelligence (AI) and natural language processing (NLP)
have improved dramatically.
Logical reasoning technology has advanced while production rule
technology remains stagnant” (Paul Haley 2018)
• http://haleyai.com/wordpress/2018/04/02/confessions-of-a-production-rule-vendor-part-2/

58. Still Here
58
Working MemoryModules
Inference Engine
Pattern Matching
Agenda

59. Still Here
59

60. Drools areas of
Improvements
60

61. •Pluggable Knowledge Types
•Canonical Model
•Separation of rules from data
•Improved, declarative, rule execution control (Instead of push/pop stacks)
•Common “unit of execution” for runtimes (rules, workflow, bpm).
•Improved APIs and programming model
•Pluggable Belief Systems
Areas of Improvements
61

62. Simple TMS
62
•A rule “logically” inserts an object
•When the rule is no longer true, the object is retracted.
rule "IsChild"
when
$p : Person( age < 16 )
then
logicalInsert( new IsChild( $p ) )
end
rule "IsAdult"
when
$p : Person( age >= 16 )
then
logicalInsert( new IsAdult( $p ) )
end
rule "Issue Child Bus Pass"
when
$p : Person( )
IsChild( person =$p )
then
logicalInsert(new ChildBusPass( $p ) );
end
rule "Issue Adult Bus Pass"
when
$p : Person()
IsAdult( person =$p )
then
logicalInsert(new AdultBusPass( $p ) );
end

63. Defeasible
63
rule "All Birds Fly" @Defeasible
when
$b : Bird( )
then
logicalInsert(new Fly( $b ) );
end
rule "Penguins With Rockets Fly”
@Defeasible @Defeats(“Penguins Don't Fly”)
when
$b : Penguin( )
Rocket($b;)
then
logicalInsert(new Fly( $b ) );
end
rule "Penguins Don't Fly"
@Defeasible @Defeater
when
$b : Bird( )
Penguin($b;)
then
logicalInsert(new Fly( $b ), “neg” );
end

64. Bayesian Network
64

65. Bayesian Network
65
KnowledgeBuilder kBuilder = KnowledgeBuilderFactory.newKnowledgeBuilder();
kBuilder.add( ResourceFactory.newClassPathResource(“rules.drl”), ResourceType.DRL );
kBuilder.add( ResourceFactory.newClassPathResource(“Garden.xmlbif"), ResourceType.BAYES );
KnowledgeBase kBase = KnowledgeBaseFactory.newKnowledgeBase();
kBase.addKnowledgePackages( kBuilder.getKnowledgePackages() );

66. Pluggable Beliefs - Bayesian Network
66
BayesInstance<Garden> instance = bayesRuntime.createInstance(Garden.class);
BayesVariable var = ( BayesVariable ) instance.getFieldNames().get( “WetGrass” );
bayesInstance.setDecided(var, true);
bayesInstance.setLikelyhood( var, [1.0,0.0] );
Garden garden = instance.marginalize();;
FactHandle fh = ksession.insert( garden );
ksession.insert( new Cloud() );
ksession.fireAllRules();
rule "I see clouds" when
g : Garden(decided == false)
exists Cloud()
then
insertLogical( bayesVar(g, 'Cloudy'),
evidenceFactory.create( new double[] {1.0,0.0} ) );
end
rule "It probably rained" when
Garden(decided == true, rain == true)
then
System.out.println( "Rain" );
end

67. Explainable AI (XAI)
67

68. DARPA : Explainable AI (XAI)
Explainability / Trustworthiness / Compliance