Engineering Data-Centric Process Models · Data-centric processes . Data-centric process model combines data and behavior perspective • Artifact-driven BPM (IBM) • Case handling/management

Engineering Data-Centric Process Models

Rik Eshuis

Process modeling perspectives

PAGE 2 22-10-2014

Behavior view

Data view

Resource view

Business Process

Separation of concerns: engineer models separately

Information System

Data-centric processes

Data-centric process model combines data and behavior perspective • Artifact-driven BPM (IBM) • Case handling/management • Document-driven workflows (PSU) • Data-driven process structures (Ulm Univ) • Process models with data flows

PAGE 3 22-10-2014

Why data-centric processes?

• Gaining flexibility

PAGE 4 22-10-2014

classic

data-centric

Why data-centric processes?

• Gaining flexibility • Gaining precision • Gaining understandability

− for business users, thinking in terms of entities (artifacts) is more easy then thinking in terms of processes [IBM]

PAGE 5 22-10-2014

Data-centric processes

Data-centric process model combines data and behavior perspective • Artifact-driven BPM (IBM) • Case handling/management • Document-driven workflows (PSU) • Data-driven process structures (Ulm Univ) • Process models with data flows

PAGE 6 22-10-2014

Outline

• Framework • Analysis of workflow data models • Synthesis of object-life cycles • Transactional process views • Splitting GSM schemas

PAGE 8 22-10-2014 22-10-2014 PAGE 8

Control flow and data flow

PAGE 9 22-10-2014

A B C

message passing: loose integration

shared database: tight integration

Data granularity

• ~ how many states can a data item have? • Variable: each value is a state • Object: discrete, explicitly specified states

PAGE 10 22-10-2014

variables objects

low granularity high granularity

Business process scope

• Intra-organizational • Inter-organizational

PAGE 11 22-10-2014 22-10-2014 PAGE 11

Data flow dimensions

• Data integration − Tight: Shared database − Loose: Message passing

• Data granularity

− High: variables − Low: object states

• Scope

− Intra-organizational − Inter-organizational

PAGE 12 22-10-2014

Shared database Message

Object Variable Object Variable Intra-organizational Inter-organizational

Research topics

• Analysis of workflow data models • Synthesis of object-life cycles • Transactional process views • Splitting GSM schemas

PAGE 13 22-10-2014 22-10-2014 PAGE 13


Object Variable Object Variable Intra-organizational

Analysis of workflow data models

Synthesis of object-life cycles

Inter-organizational Transactional process views

Splitting GSM schemas

Research topics


PAGE 14 22-10-2014







22-10-2014 PAGE 14

Analysis of workflow data models (joint work with Diana Borrego of Un. of Seville)

• Workflow models with variables and pre/postconditions for activities

• Workflow gets stuck at activity if precondition fails • How can we diagnose correctness of workflow

models with pre/postconditions?

22-10-2014 PAGE 15

Example

Precondition

violation

22-10-2014 PAGE 16

Diagnosis

3*30+100+30<200

22-10-2014 PAGE 17

Diagnosis approach

• Workflow graphs + pre/postconditions • Encode correct control flow execution (subgraph) as

Integer Programming model − Rik Eshuis, Akhil Kumar: An integer programming based approach for verification

and diagnosis of workflows. Data & Knowl. Eng. 69(8): 816-835 (2010)

• Encode pre/postconditions as Constraint Satisfaction Problem

• Combined IP+CSP model processed by CSP solver

22-10-2014 PAGE 18

Algorithm

• For each activity a with precondition p • Find subgraph that leads to a but does not include a • Encode subgraph as CSP+IP model • Add not(p) to CSP+IP model • If solution exists, then error

• If no error found, then the workflow graph is correct

PAGE 19 22-10-2014

Validation

• Java-based implementation for verifiying annotated XPDL models

• Using COMET CSP checker • The execution time scales linearly in the size of the

XPDL models

PAGE 20 22-10-2014

Results

Contributions • Seamless integration of process and data

verification • Extensive feedback to fix data flow errors • Fully implemented

Future work • Extend set of workflow operators

22-10-2014 PAGE 21

Diana Borrego, Rik Eshuis, María Teresa Gómez López, Rafael M. Gasca. Diagnosing correctness of semantic workflow models. Data & Knowledge Engineering, 87: 167-184 (2013)

Research topics


PAGE 22 22-10-2014







22-10-2014 PAGE 22

Synthesis of object life cycles (joint work with Pieter Van Gorp)

• UML activity diagrams implicitly specify object life cycles

• Object life cycles are specified in UML statecharts • How can we discover implicit object life cyles from

UML activity diagrams? • Why?

− Check consistency with pre-existing object life cycles − Generate software code

Withdraw cash

Check balance Dispense cash

Log entryPrint receipt

: Account[uncredited]

: Account[checked]

: Account[credited]

22-10-2014 PAGE 23

Receive

Reject

Determine costs

Update client contributionNotify client

Pay

: Claim[policy checked]

: Claim[rejected]

: Claim[archived]

: Claim[policy not checked]

: Claim[damage checked] : Claim

[settled]

Check policy

Check damage

: Claim[assessed]

: Claim[damage not checked]

: Claim[received]

received

rejected

assessed

policy checked

damage checked

archived

settled

checking

O4

O3

policy not checked

damage not checked

O1

O2

preprocessing

???

22-10-2014 PAGE 24

Approach

Two sequential steps • Filtering (to remove irrelevant parts) • Synthesis (to construct statechart hierarchy) Fully automated!

PAGE 25 22-10-2014

Filtering

• Reduction rules

22-10-2014 PAGE 26

Filtering effect

: Claim[policy checked]

: Claim[rejected]

: Claim[archived]

: Claim[policy not checked]

: Claim[damage checked]

: Claim[settled]

: Claim[assessed] : Claim

[damage not checked]

: Claim[received] : Order[accepted]

: Order[accepted]

: Order[accepted]

: Order[accepted]

22-10-2014 PAGE 27

Synthesis

• Transformation rules to construct state hierarchy − rule for creating AND state − rule for creating OR state

22-10-2014 PAGE 28

Example construction AND state

22-10-2014 PAGE 29

received

O7

O2

O1

received

damage not checked damage checked

policy not checked policy checkedO7

O2

O35

O46

O1

T2

damage not checked damage checked

policy not checked policy checkedO35

O46

A1O3456

Final statechart

• Add transitions to statechart hierarchy

received

rejected

assessed

policy checked

damage checked

archived

settled

checking

O4

O3

policy not checked

damage not checked

O1

O2

preprocessing

22-10-2014 PAGE 30

Implementation

PAGE 31 22-10-2014

GrGenGraph Rewriting Tool

with optional Visualization

Eclipse based editors(Eugenia/Kieler and Official UML 2 Tools)

UML Activity Diagram

XMI

UML Statechart

XMI

Filteringrules

Synthesisrules

Final remarks

Contributions • Synthesis of expressive hierarchical state machines • Structure preserving • Filtering out of irrelevant parts • Fully implemented prototype using GrGen • Extended to deal with multiple objects Future work • Enhance set of translations • Support roundtrip engineering

22-10-2014 PAGE 32

Rik Eshuis, Pieter Van Gorp. Synthesizing Object Life Cycles from Business Process Models. Proc. ER 2012, Springer (2012). Rik Eshuis, Pieter Van Gorp. Synthesizing Object-Centric Models from Business Process Models. Proc. BPM 2013 workshops, Springer (2014).

Research topics


PAGE 33 22-10-2014







Transactional Process Views (Joint work with Jochem Vonk & Paul Grefen)

• Process view reveals relevant properties of underlying private business process

• Balance between interest of owner and client

• How to construct a transactional process view from an internal transactional process?

No transactional

support

22-10-2014 PAGE 34

Example classical process view

Classic process views

R. Eshuis and P. Grefen. Constructing Customized Process Views. Data and Knowledge Engineering, 64(2):419-438 (2008).

Example internal transactional process

PAGE 37 22-10-2014

Approach

• Extend process view construction approach with transactional features − Aggregation − Customization

???

22-10-2014 PAGE 38

Final remarks

Contribution • Flexible construction of transactional process views • Specification of robust & reliable behaviour at the

external level • Approach helps to establish dynamic trustworthy

B2B collaborations in an efficient way Future topics: • More expressive process models • Execution of transactional process views

PAGE 39 22-10-2014

Rik Eshuis, Jochem Vonk, Paul Grefen. Transactional Process Views. Proc. OTM Conferences (1) 2011, pp.119-136 (2011)

Research topics


PAGE 40 22-10-2014







22-10-2014 PAGE 40

Guard-Stage-Milestone (GSM) schemas

• Status of each stage,milestone ∈ { on, off} • Business rules specify when stage/milestone changes state

− Might refer to status of other stages/milestones • Rules need to be evaluated in right order to ensure that all

changes have maximal effect • Unit of change triggered by external event is called a B-step

22-10-2014

DealRefined

DealFailed

Refine Deal

Term DraftedDraft Term Credit

CheckedCheck CreditDetermine

PricePrice

Determined

ReDraftTermNeeded ReCheckCreditNeeded

Guard business rule when

to open stage

Stage cluster of work to achieve milestone

Milestone business objective

Splitting GSM subschema (Joint work R. Hull, Y. Sun, R. Vaculin from IBM Research)

22-10-2014 PAGE 42

Client artifact system

Provider artifact system

must be equivalent

Deal Refined

Deal Failed

Refine Deal

Term Drafted Draft Term Credit

Checked Check Credit Determine

Price Price

Determined

ReDraftTermNeeded ReCheckCreditNeeded

Draft Term Deal

Refined

Deal Failed

Refine Deal

Credit Checked Check Credit

Determine Price

Price Determined

ReCheckCreditNeeded ReDraftTermNeeded

IGF

LawOffice IGF

Term Drafted

Naïve splitting causes race conditions

22-10-2014 PAGE 43

Deal Refined

Deal Failed

Refine Deal

Credit Checked Check Credit Determine

Price Price

Determined

ReCheckCreditNeeded

Term Drafted

ReDraftTermNeeded

Draft Term achieve if credit_level > 100,000

open on +ReDraftTermNeeded

open if TermDrafted and CreditChecked status is on status is off

Task CheckCredit completes

credit_level= 200,000

IGF

LawOffice

Naïve splitting causes race conditions

22-10-2014 PAGE 44

Deal Refined

Deal Failed

Refine Deal

Credit Checked Check Credit Determine

Price Price

Determined

ReCheckCreditNeeded

Term Drafted

ReDraftTermNeeded

Draft Term achieve if credit_level > 100,000

open on +ReDraftTermNeeded

open if TermDrafted and CreditChecked status is on status is off

Task CheckCredit completes

credit_level= 200,000

IGF

LawOffice

Impossible in original GSM

schema

Possible solutions

1. Extend local GSM engines with complex synchronization logic

2. Better: restructure GSM schema to avoid complex synchronization

22-10-2014 PAGE 45

Event B-step 1

Event’ B-step 2

IGF

LawOffice

Solution approach

• Keep local cache of status of needed remote stages/milestone • determined per external event type

• Reallocate business rules to avoid complex synchronization − Sometimes: rule executes at different location than

the corresponding stage/milestone • Runtime simulation protocol based on distributed 2-

phase commit • Result (see paper): split GSM schema faithfully

simulates original GSM schema

22-10-2014 PAGE 46

Final remarks

• Comprehensive framework for supporting outsourcing of GSM schemas

• Covers both design-time and run-time • Future work:

• multi-party outsourcing • interacting artifact types • extend results to OMG’s Case Management Modeling

Notation

22-10-2014 PAGE 47

Rik Eshuis, Richard Hull, Yutian Sun, Roman Vaculín. Splitting GSM Schemas: A Framework for Outsourcing of Declarative Artifact Systems. Proc. BPM 2013, pp. 259-274 (2013)

Conclusion

Future work • Inter-organizational perspective • Case management







22-10-2014 PAGE 48

For more info …

Contact details • Email: [email protected] • Homepage: http://is.ieis.tue.nl/staff/heshuis

PAGE 49 22-10-2014

Thank you!

Documents

Engineering Data-Centric Process Models · Data-centric processes . Data-centric process model combines data and behavior perspective • Artifact-driven BPM (IBM) • Case handling/management