Lazy Soundness - frapu.de · Loan Broker Bank Buy direct Find Bank Request Loan Buy Reject Purchase...
Transcript of Lazy Soundness - frapu.de · Loan Broker Bank Buy direct Find Bank Request Loan Buy Reject Purchase...
Anwendung einer Theorie für Mobile Systeme auf Geschäftsprozessmanagement
Frank Puhlmann
Fachgebiet Business Process TechnologyHasso Plattner InstitutPotsdam
On the Application of a Theory for Mobile Systems to Business Process Management
Lazy SoundnessA Prototypical Tool-Chain
Lazy Soundness is a new kind of soundness dealing with so calledleft-behind or lazy activities. Since these activities can be active while the final activity of the business process has already been reached, processes containing these activities can never be sound. Lazy soundness provides a criterion to prove business processes containing these activities to be free of deadlocks and livelocks.
Prof. Dr. Mathias WeskeFrank PuhlmannBusiness Process Technology GroupHasso Plattner InstituteCampus Griebnitzsee14482 Potsdam, Germany
http://bpt.hpi.uni-potsdam.de
A business process containing Discriminator, N-out-of-M, or Multiple Instances without Synchronization patterns (called the critical patterns), such as
A
B
C
2 D
Problem
Solution
Structural
Sound Process
Initial
Node
Final
Node
A, B, and C represent three web service interactions.
After two of them have completed, D is executed and thereafter the process is finished.
However, one of the activities is still active, and clean-up work like payment and documentation has to be done.
As the remaining activity contradicts the soundness definition, we can not use existing tools to verify the sample business process. Still, automated verification regarding deadlocks and livelocks is quite important even if you employ one of the critical patterns in your business process.
Lazy Soundness proves business processes containing the critical patterns (and all others) to be free of deadlocks and livelocks. Technically, it abstracts from all internals of the process and just considers the initial and final node. The abstracted process is verified using bisimulation techniques.
Demo Presentation:
Thursday, Sep 7 10:30am, Room EI10
Lazy soundness has been implemented in a prototypical tool chain at our research group. We provide a graphical editing of business processes using BPMN, automatically formalize BPM diagrams into pi-calculus expressions, and use existing tools to decide lazy soundness for a given business process.
The theoretical background of Lazy Soundness will be presented on Tuesday, September 5 16:30am, Room EI9.
Grundlagen
Shop
Dienste
Send
Order
Receive
Order
Custo
mer
Geschäftsprozess-management
Schlüsselkonzepte
Receive Order
Process Order
Pack Item
Create Invoice
Ship Order
Receive Payment
More Items?
Aktivitäten
BeziehungenSales
Distribution
Fulfillment
Accounting
Rollen
ORDER
ORDER
ORDER
INVOICE
SHIPMENT
SHIPMENT
INVOICE
Daten
Interaktionen
3
Design & Analyse
Konfiguration
Abwicklung
Evaluierung
Geschäftsprozess-management
Lebenszyklus
4
Thema der Arbeit Design & Analyse
Design & Analyse
Graphischer Prozess/
Interaktion
Validierung
Pi-Kalkül Darstellung
Abwickelbare Darstellung
SimulationVerifizierung
5
Das Pi-Kalkül Mobilität
6
R S
S
S
S
S
S
B
Beitrag
Ndef= (!e1 . . . e8 )
7!
i=1
Ni
Prozesse Formale Darstellung
8
Receive Order Process
Order
Pack Item
Create Invoice
Ship Order
Receive Payment
More Items?
N1
N2
N3 N4
N6
N5
N7
e1
e2
e3
e4e6
e7
e8
e5
Prozesse Muster
9
Create
Invoicea
b Adef= a.!·".b.0
Sequence
Process
Order
a
b1
b2
Adef= a.!·".(
n!
i=1
bi.0)
Parallel Split
Ship Order
a1
a2
b
Adef= {ai}
n
i=1.!·".b.0
Synchronization
Process Graph
InitialNode
FinalNode
Prozesse Verifikation
10
Lazy Soundness: Ein wohlgeformter Prozessgraph ist lazy sound, wenn in jedem Fall der finale Knoten genau einmal erreicht wird.
Receive Order Process
Order
Pack Item
Create Invoice
Ship Order
Receive Payment
More Items?
N1
N2
N3 N4
N6
N5
N7
e1
e2
e3
e4e6
e7
e8
e5S
ho
p
Send
Order
Receive
Order
M1 M2
f1
Cu
sto
me
r
s(r) r
Idef= (N | M | . . . )
Interaktionen Formale Darstellung
11
Send Order
s(r)
Send
Receive
Receive
Order
r
Interaktionen Muster
12
Idef= (!s)(s!r".N ! | s(r).M ! | . . . )
a) statisch
Idef= (!get)(get(s).s!r".N ! | . . . )
b) dynamisch
Loan B
roker
Bank
Buy direct
Find Bank
Request Loan
Buy
Reject Purchase
v< !1.000
v> !999
Lookup Banks
Send Accept
Accept
Reject
Send Reject
Custo
mer
B1 B2 B3b1 b2
S1S2
S3
S4
S5S6
s1
s2
s3
s4
s5
s6
broker(ch) [1] ch(bank) [2] bank(req,acc,rej) acc rej
C1C2
C3
C4
C5
C6
C7
C8
C9
C10
C11
C12c1
c2
c3
c4
c5
c6
c7
c8
c9
c10
c11
c12 c13
Interaktionen Verifikation
13
Interaction Soundness: Ein Dienst ist interaction sound in Bezug auf eine Umgebung, wenn die Vereinigung aus Dienst und Umgebung lazy sound ist.
Ergebnisse
Diskussion des Pi-Kalküls als formale Grundlage
• Vorteile
• Direkte Darstellung von dynamischem Binden
• Unterstützung von Kompositionalität als ing.-mäßige Vorgehensweise
• Unterstützung aller relevanten Konzepte
• Elegante Repräsentation von Invarianten zur Verifizierung
• „Prototypische“ Sichtweise
15
• Nachteile
• Komplexe, mathematische Darstellung
• Hohe Aufwände für Bisimulationsprüfung
• Bisimulation nicht für Konformitätsprüfungen von Diensten geeignet
• Transitionssemantik (keine „Erzwingung“ einer Transition)
• „Prototypische“ Sichtweise
• Einheitliche, formale Sicht auf Daten, Prozesse und Interaktionen
• Formale Darstellung der dazugehörigen Pattern
• Algorithmen zur Abbildung von Prozess-/Interaktiongraphen auf das Pi-Kalkül
• BPMN Erweiterungen für dynamisches Binden
• Lazy Soundness
• Interaction Soundness
• Prototypische Werkzeugkette
Zusammenfassung
• Grundlagen
• Geschäftsprozessmanagement
• Pi-Kalkül
• Beitrag
• (Daten)
• Prozesse
• Interaktionen
• Ergebnisse
• Diskussion
• Zusammenfassung
16
Beitrag
Publications based on this Thesis
Early ideas have been published and presented at national and international conferences duringthe writing of this thesis. A starting point was a conference paper at the third conference onbusiness process management (BPM) in Nancy (France).1 It showed how the !-calculus mightbe used to represent the workflow patterns. With this paper, a first draft of the pattern formal-izations contained in chapter 5 (Processes) has been brought to a larger audience. While theformalizations in most cases did not required advanced features of the !-calculus, such as linkpassing mobility, a subsequent paper revealed the strengths of the !-calculus for representingdynamic binding and correlation handling in service-oriented architectures.2 It laid the foun-dations for section 6.1.1 (Correlations and Dynamic Binding). This paper has been presentedat a workshop covering dynamic web processes alongside the third international conference onservice-oriented computing (ICSOC) held in Amsterdam (The Netherlands). The investigationcontinued with a conference paper that discussed shifting requirements for BPM.3 Beside theinvestigation of state-of-the-art, new requirements regarding technical and theoretical founda-tions have been found. Refined versions of these requirements are used to motivate the thesisin chapter 1 (The Shifting Focus). The results have been presented at the ninth conference onbusiness information systems (BIS) in Klagenfurt (Austria). Thereafter the research focusedon soundness properties of business processes formalized in the !-calculus. In contrast to ex-isting properties, the application of bisimulation equivalence for reasoning on deadlock andlivelock freedom has been investigated. Based on an extensive study of the workflow patternformalizations, it turned out that several of them constitute problems regarding soundness. Theproblems have been overcome by a new soundness property that was named lazy soundness.A refined version is contained in section 5.3 (Properties) of chapter 5 (Processes). The newsoundness property has been presented to the scientific community at the fourth conference onbusiness process management (BPM) in Vienna (Austria), where it has been published as part
1 Frank Puhlmann, Mathias Weske: Using the Pi-Calculus for Formalizing Workflow Patterns. In W.M.P. van derAalst, B. Benatallah, F. Casati, and F. Curbera (Eds.): Business Process Management, volume 3649 of LNCS,Nancy, France, Springer-Verlag (2005) 153–168
2 Hagen Overdick, Frank Puhlmann, Mathias Weske: Towards a Formal Model for Agile Service Discovery andIntegration. In K. Verma, A. Sheth, M. Zaremba, and C. Bussler (Eds.): Proceedings of the International Work-shop in Dynamic Web Processes (DWP 2005), Amsterdam, The Netherlands, IBM technical report RC23822(2005)
3 Frank Puhlmann: Why do we actually need the Pi-Calculus for Business Process Management? In W.Abramowicz and H. Mayr (Eds.): BIS 2006—Business Information Systems, volume P-85 of LNI, Klagenfurt,Austria, Gesellschaft fuer Informatik (2006) 77–89
iii
iv
of the conference proceedings.4 The practical feasibility of lazy soundness has been shownin an additional presentation, published as part of the demo session proceedings.5 At the sameconference, a short paper, written together with a student of mine, gave an insight on how the ser-vice interaction patterns might be formalized in the !-calculus.6 While chapter 6 (Interactions)contains a different approach for representing these patterns, the paper nevertheless providedvaluable ideas. The publication series continued with a paper presented at a national conferenceon service-oriented information systems (EMISA) that took place in Hamburg (Germany).7 Itcovered the unification of data, processes, and interactions to provide a unified formal represen-tation of service-oriented architectures. The discussion has been based on an example that canbe found in an extended version in chapter 7 (Unification). In the meantime, a book chapter onthe suitability of the !-calculus for BPM has been published.8 It basically contains an extendedand updated version of the BIS paper published earlier. Another publication covers an exten-sion of lazy soundness to prove compatibility in interactions.9 The new compatibility property,denoted as interaction soundness, supports dynamic binding. To the knowledge of the author,this was the first paper that introduced compatibility with dynamic binding. The updated resultscan be found in section 6.3 (Interaction Soundness) of chapter 6 (Interactions). The paper hasbeen presented at the fourth international conference on service-oriented computing (ICSOC)in Chicago (USA). Furthermore, during the writing of this thesis, the author supervised a mas-ter thesis where a graphical environment for the simulation of business processes with dynamicbinding—based on the ideas found in the second and third part of this work—has been imple-mented. The corresponding tool has been presented at the open.BPM workshop in Hamburg(Germany).10
4 Frank Puhlmann, Mathias Weske: Investigations on Soundness Regarding Lazy Activities. In S. Dustdar, J.L.Fiadeiro and A. Sheth (Eds.): Business Process Management, volume 4102 of LNCS, Vienna, Austria, Springer-Verlag (2006) 145–160
5 Frank Puhlmann: A Tool Chain for Lazy Soundness. Demo Session of the 4th International Conference onBusiness Process Management, CEUR Workshop Proceedings Vol. 203, Vienna, Austria (2006) 9–16
6 Gero Decker, Frank Puhlmann, Mathias Weske: Formalizing Service Interactions. In S. Dustdar, J.L. Fiadeiroand A. Sheth (Eds.): Business Process Management, volume 4102 of LNCS, Vienna, Austria, Springer-Verlag(2006) 414–419
7 Frank Puhlmann: A Unified Formal Foundation for Service Oriented Architectures. In M. Weske and M.Nuettgens (Eds.): EMISA 2006, volume P-95 of LNI, Hamburg, Germany (2006) 7–19
8 Frank Puhlmann: On the Suitability of the Pi-Calculus for Business Process Management. In Technologies forBusiness Information Systems. Springer-Verlag (2007) 51–62
9 Frank Puhlmann, Mathias Weske: Interaction Soundness for Service Orchestrations. In A. Dan and W. Lamers-dorf (Eds.): Service-Oriented Computing, volume 4294 of LNCS, Chicago, USA, Springer-Verlag (2006) 302–313
10 Anja Bog, Frank Puhlmann: A Tool for the Simulation of Pi-Calculus Systems. In 1. GI-Workshop OpenBPM2006: Geschaftsprozessmanagement mit Open Source-Technologien, Hamburg, Germany (2006)
Publikationen Zur Dissertation
17
Vielen Dank für Ihre Aufmerksamkeit!