Composite Event Specification in Active Database Systems: A Petri Nets Approach Xiaoou Li, Joselito Medina Marín, Lorena Chavarría Petri Nets Research.

Composite Event Specification in Active Database Systems: A Petri Nets Approach

Xiaoou Li,

Joselito Medina Marín,

Lorena Chavarría

Petri Nets Research Group

CPN'04, Aarhus, 10/10/2004 2

Sección de ComputaciónDepartamento de Ingeniería Eléctrica Centro de Investigación y de Estudios Avanzados del IPN MEXICO

http://www.cs.cinvestav.mx/

CPN'04, Aarhus, 10/10/2004 3

Outline:

Motivation Event-Condition-Action (ECA) rules Conditional Colored Petri Nets (CCPN) ECA-PN Simulator (ECAPNSim) Termination analysis Complexity analysis Conclusion

CPN'04, Aarhus, 10/10/2004 4

Motivation

OurOur experience: experience: X. Li, W. Yu and F. Lara, “Dynamic X. Li, W. Yu and F. Lara, “Dynamic knowledge inference and learning under knowledge inference and learning under adaptive fuzzy Petri net framework”, adaptive fuzzy Petri net framework”, IEEE Trans. SMC-CIEEE Trans. SMC-C, 30(4), 442-450, 2000, 30(4), 442-450, 2000

CCPNCCPN

ECAPNSimECAPNSim Applications

• The knowledge and execution The knowledge and execution models of ECA rules are models of ECA rules are separated;separated;• Rule base management is Rule base management is application-dependent.application-dependent.

CPN'04, Aarhus, 10/10/2004 5

ECA rules

ON event

IF condition

THEN action

ON update BONUS_amount, IF update.amount > 100, THEN update EMP set rank = old.rank + 1

where emp_id = update.emp_id;

CPN'04, Aarhus, 10/10/2004 6

ECA rule execution

User Transactions DB

DBAction

Condition over

DB State

C-A of Rule

Other Actions

Event Detector

Fire RuleEvent

Other Events

Action

CPN'04, Aarhus, 10/10/2004 7

Events in ECA rules

Primitive events Composite events

1. Conjunction: e1 e22. Disjunction: e1 e23. Sequence: seq(e1, e2)4. Simultaneous: sim (e1, e2)5. History: times(n, e1) in [d1, d2]6. Negation: ~e1 in [d1, d2]7. Any: any(m, e1, e2, …, en), n>m

CPN'04, Aarhus, 10/10/2004 8

Conditional Colored Petri Nets-CCPN

CCPN={Σ, P, T, A, N, C, Con, Action, D}

Con is a condition function. ∀t T∈ rule: Type(Con(t))=B, where Con function evaluates the

rule condition; ∀t T∈ comp: Type(Con(t))=B, where Con function evaluates the

temporal condition.

D assigns a time interval [d1, d2] to t T∈ comp.

Action is an action function. ∀t T∈ rule, p∈ : Type(Action(t))=C(p)MS.

Token element TE (p,c,timestemp)

t

CPN'04, Aarhus, 10/10/2004 9

CCPN elements

Composite transition

Rule transition

Primitive place

Copy place

Virtual place

Composite place

Copy transition

Normal arc

Inhibitor arc

CPN'04, Aarhus, 10/10/2004 10

Modeling ECA rules with CCPN

ON event

IF condition

THEN action

T1 RULE 1

CPN'04, Aarhus, 10/10/2004 11

Modeling ECA rules with CCPN

rule 001,ON update BONUS_amount, IF update.amount > 100, THEN update EMP set rank = old.rank + 1

where emp_id = update.emp_id;

p1

p2

T1p1: update_BONUS_amountT1: rule 001condition on T1: update.amount > 100p2: update_EMP_rank token color in p1: any BONUS_amounttoken color in p2: EMP_rank = old.rank + 1

where emp_id = update.emp_id

CPN'04, Aarhus, 10/10/2004 12

Modeling Composite events with CCPN…

Conjunction Disjuntion

Color ec : e1 e2

e1 e2

ec

T1

e1 e2

ec

Color ec : e1 e2

T1 T2

CPN'04, Aarhus, 10/10/2004 13


Negation

Color ec : ~e1 in [d1, d2]

e1

ec

T1

Sequence

Color ec : seq(e1, e2)

e1 e2

ec

T1 if (e1) < (e2)

CPN'04, Aarhus, 10/10/2004 14


Simultaneous

Color ec : sim (e1, e2)

e1 e2

ec

if (e1) = (e2) T1

Color ec : times(n, e1) in [d1, d2]

e1

ec

n

T1

History

CPN'04, Aarhus, 10/10/2004 15


Any

Color ec : any(m, e1, e2, …, en)

e0

m

e1 e2 en…

…

ec

T2T1 Tn

T0

CPN'04, Aarhus, 10/10/2004 16

Conversion from ECA rules to CCPN

Algorithm:

Input:X.eca

Output: CCPN

CPN'04, Aarhus, 10/10/2004 17

Modeling ECA rules with composite events…

Tables

EMPLOYEE (emp_id, name, rank, salary); SALES (emp_id, date, sale);

Rulesrule 001, ON or(e0:e1), IF update.salary > 50000, THEN update EMPLOYEE set salary

= update.salary * 0.20 where

emp_id = update.emp_id;rule 002, ON and(e0:e2), IF insert.number > 10,THEN update EMPLOYEE set salary

= insert.salary * 1.10 where

emp_id = insert.emp_id;

Events

e0 : insert_EMPLOYEE; e1 : update_EMPLOYEE_salary; e2 : insert_SALES;

CPN'04, Aarhus, 10/10/2004 18

Modeling ECA rules with composite events…

CPN'04, Aarhus, 10/10/2004 19

Usefulness of CCPN

ECA rule execution simulation and supervision Static and dynamic property analysis ECA rule interaction complexity analysis An engine of an active database system

CPN'04, Aarhus, 10/10/2004 20

Database

Composite Event Detector

ECAPNSimTransactions

CCPN Manager

ECA ruleeditor

Terminationanalyzer

ECA - CCPNconvertor

Explanationcomponent

Tools Environment

CCPN editor/visualizer

ECA rules developer

ECAPNSim Kernel

CCPNBase

Console

Rule ExecutionComponent

event signalling CCPN retrieval event rule execution

User

ECAPNSim architecture

Complexityanalyzer

CPN'04, Aarhus, 10/10/2004 21

ECA rule Editor

ECA - CCPNConvertor

CCPNEditor / Visualizer

ECA rules developer

ECAPNSim Kernel

CCPNbase

Termination Analyzer – The Engine

Incidence matrix generator

Paths Searcher

Paths Analyzer

Cyclic Paths Detector

RefinedCycles

CCPNManager

ECA rule Compiler

ECA rules developer

ECAbase

Static tools

Data flow

CCPN elements retrieval and storage

Explanation component

ECAPNSim Buildtime Tools

CPN'04, Aarhus, 10/10/2004 22

Rule Execution Component

DB

Event Detector

CCPN Manager

Simulation mode

CCPN editor/

visualizer

User

Real mode

CCPNbase

Console

Dataflow at runtime

ConditionEvaluator

ActionExecutor

With token information

Random evaluation

Condition evaluator

Token gameanimation

ECAPNSim Kernel Static tools

ECA rules developer

ECAPNSim

ECAPNSim Runtime Tools

CPN'04, Aarhus, 10/10/2004 23

Applications of CCPN

TraditionalDB

ECAPNSim Active DB+ =

CPN'04, Aarhus, 10/10/2004 24

A realized active database system

System: Access

Relations: EMP(emp_id, name, rank, salary);

BONUS(emp_id, amount);

SALES(emp_id, month, number);

CPN'04, Aarhus, 10/10/2004 25

A realized active database system…

CPN04.ecarule 001,

ON update_BONUS_amount, IF update.amount > 100, THEN update EMP set rank = old.rank + 1 where emp_id = update.emp_id;

rule 002, ON update_EMP_rank, IF update.rank > 5, THEN update BONUS set amount = old.amount + update.rank * 10 where emp_id =

update.emp_id;rule 003,

ON insert_SALES, IF insert.number > 50, THEN update EMP set rank = old.rank+1 where emp_id = insert.emp_id;

rule 004, ON update_EMP_rank, IF update.rank = 15, THEN update EMP set salary = old.salary*1.1 where emp_id = update.emp_id;

CPN'04, Aarhus, 10/10/2004 26

Termination Analysis with CCPN

CPN'04, Aarhus, 10/10/2004 27

Termination Analysis with CCPN…

CPN'04, Aarhus, 10/10/2004 28

Complexity Analysis with CCPN

Definition of parameters: distance, potential, anchor.

Calculation of the parameters based on incidence matrix

Validation based on Zuse’s software measurement framework

Implementation Case study: 50 rules

CPN'04, Aarhus, 10/10/2004 29

Complexity Analysis…

//Definition of tablesDEP ( ElDep CHAR( length), Prod CHAR( length) , Presupuesto float ) ; TRABAJA ( ElProy CHAR( length), ElEmp CHAR( length)) ;EMP ( SuDep CHAR( length), ElEmp CHAR( length), Salario float, Bono

float ) ;PROYECTO ( ElProy CHAR( length), Presupuesto float ) ;

//Definition of eventse0 : delete_DEP ; e1 : delete_TRABAJA ; e2 : delete_EMP ; e3 : insert_EMP;e4 : update_EMP_Salario; e5 : or(e3:e4) ;e6 : update_DEP_Prod; e7 : and(e6:e3);e8 : update_DEP_Presupuesto; e9 : insert_DEP;e10 : or(e8:e9);e11 : update_PROYECTO_Presupuesto;

CPN'04, Aarhus, 10/10/2004 30

Complexity Analysis…//Definition of rulesrule 000, ON e0, IF true, THEN delete from EMP where EMP.SuDep = DEP.ElDep;rule 001, ON e1,IF true, THEN delete from PROYECTO where PROYECTO.ElProy = TRABAJA.ElProy;rule 002, ON e2, IF true, THEN delete from TRABAJA where EMP.ElEmp = TRABAJA.ElEmp;rule 003, ON e5, IF EMP.Salario > 15000, THEN delete from EMP where EMP.ElEmp = new.EMP.ElEmp;

CPN'04, Aarhus, 10/10/2004 31


//Definition of rulesrule 004, ON e7, IF DEP.Prod > 90 & EMP.SuDep = DEP.ElDep, THEN update EMP set value Bono = 100 where EMP.ElEmp = new.EMP.ElEmp;rule 005, ON e10, IF true, THEN update PROYECTO set value Presupuesto = 0.1 * PROYECTO.Presupuesto

where PROYECTO.ElProy = TRABAJA.ElProy; rule 006, ON e11, IF true, THEN update EMP set value Salario = 0.01 * PROYECTO.Presupuesto where

PROYECTO.ElProy = TRABAJA.ElProy;

CPN'04, Aarhus, 10/10/2004 32

Complexity analysis…

CPN'04, Aarhus, 10/10/2004 33


CPN'04, Aarhus, 10/10/2004 34


CPN'04, Aarhus, 10/10/2004 35


CPN'04, Aarhus, 10/10/2004 36


CPN'04, Aarhus, 10/10/2004 37


CPN'04, Aarhus, 10/10/2004 38

Conclusion

CCPN is a modified CPN, maybe it can be converted into a CPN

As an executable model, CCPN may be applied on a lot of systems where knowledge is represented by ECA rules

ECAPNSim is application-independent Composite events complexity can be measured

in CCPN framework

CPN'04, Aarhus, 10/10/2004 39

Future works

To find practice database applications to prove and improve ECAPNSim

To improve UI More work on complexity analysis

condition temporal composite events

To apply CCPN on workflow systems

Composite Event Specification in Active Database Systems: A Petri Nets Approach Xiaoou Li, Joselito Medina Marín, Lorena Chavarría Petri Nets Research.

Documents