Diagnosing and Repairing Data Anomalies in Process Models
Description
Workshop presentation given by Niels Lohmann on September 7, 2009 in Ulm, Germany at the 5th International Workshop on Business Process Design (BPD 2009); part of BPM 2009.
Transcript
Diagnosing and RepairingData Anomalies in Process ModelsAhmed AwadHassoPlattner Institute, Potsdam, GermanyGero Decker HassoPlattner Institute, Potsdam, GermanyNielsLohmann University of Rostock, Germany
Correctness of Process Models
widely accepted: soundness
no deadlocks
no livelocks
proper termination
no dead activities
These are control flow aspects!
Diagnosing and Repairing Data Anomalies in Process Models
07.09.09
Example Process: Insurance Claim Handling
sound: every claim will eventually be closed
Diagnosing and Repairing Data Anomalies in Process Models
07.09.09
Data in BPMN
Data objects
Data states (no explosion)
Object life cycles /control flow refinement
Diagnosing and Repairing Data Anomalies in Process Models
07.09.09
Process Model with Data
This model contains five deadlocks!
Diagnosing and Repairing Data Anomalies in Process Models
07.09.09
Classes of Errors
too restrictive preconditions (close and fraudulent claims)
implicit routing (XOR vs. fraud evaluation)
implicit execution order (pay vs. file)
Diagnosing and Repairing Data Anomalies in Process Models
07.09.09
Outline
control flow + data flow = interesting problems ✔
formalization BPMN’s data aspects
detection, diagnosing, and repairing of data anomalies
Diagnosing and Repairing Data Anomalies in Process Models
07.09.09
2
FormalizationBPMN’s dataaspects
BPMN and Petri nets
BPMN:
a graphical notion
support of concurrency
Petri nets:
a graphical notion
support of concurrency
formal foundation
broad tool support
Diagnosing and Repairing Data Anomalies in Process Models
07.09.09
Dijkman et al. definePetri net semanticsfor BPMN’s control flow
Petri net formalization (control flow)
pattern-based translation
complete example (control flow):
analysis tools can check soundness
Diagnosing and Repairing Data Anomalies in Process Models
07.09.09
BPMN and Petri nets (2)
BPMN:
a graphical notion
support of concurrency
several aspects in one model
Petri nets:
a graphical notion
support of concurrency
formal foundation
broad tool support
simple composition notions
Diagnosing and Repairing Data Anomalies in Process Models
07.09.09
Formalization of BPMN’s data objects
changing a state
reading a state
changing to several possible states
Diagnosing and Repairing Data Anomalies in Process Models
07.09.09
Data flow models
data flow models for settlement and claim data object
control flow model
Diagnosing and Repairing Data Anomalies in Process Models
07.09.09
Bringing it all together
synchronization of data flow and control flow by transition fusion
Diagnosing and Repairing Data Anomalies in Process Models
07.09.09
3
Detection,diagnosing,and repair ofdata anomalies
Detection of data anomalies
standard soundness checker (Woflan, LoLA) will find deadlocks
provides counterexample (= trace)
does not differentiate data flow and control flow
gives no diagnosis/repair information
Diagnosing and Repairing Data Anomalies in Process Models
07.09.09
Diagnosis data anomalies
exploit information on model:
control flow is sound
place models either control flow or data flow
each data object can only be in one state
Diagnosing and Repairing Data Anomalies in Process Models
07.09.09
Diagnosing and fixing too restrictive preconditions
Problem: if data is set to [a], activity B is disabled
Diagnosing and Repairing Data Anomalies in Process Models
07.09.09
Diagnosing and fixing too restrictive preconditions
control flow is sound
deadlock in composite model: missing data tokens
for each deadlock:
determine missing data tokens
change model
data tokens are present, or
drop data dependency
Diagnosing and Repairing Data Anomalies in Process Models
07.09.09
Diagnosing and fixing implicit routing
decision [b] vs. [c] has to be synchronized with XOR-split
Diagnosing and Repairing Data Anomalies in Process Models
07.09.09
Diagnosing and fixing implicit routing
partition state space with respect to data states
if a decision inside a partition leads to a deadlock, this decision is “unsynchronized”
synchronize decisions according to data
Diagnosing and Repairing Data Anomalies in Process Models
07.09.09
Diagnosing and fixing implicit execution order
transitions A and B are in concurrent the control flow model,but share data place
Diagnosing and Repairing Data Anomalies in Process Models
07.09.09
Take home points
data objects can introduce errors to a model
Petri nets allow for compositional models of data and control flow
data anomalies can be detected, diagnosed,and (sometimes) automatically fixed
Future work:
automated mapping back to the BPMN model
Diagnosing and Repairing Data Anomalies in Process Models
07.09.09
Final slide
Thank you for your attention!
Your todos:
discuss with me
talk to Ahmed and Gero
attend the soundness talk(Thursday, after the keynote)
get the slides athttp://slideshare.net/correctsystems
Diagnosing and Repairing Data Anomalies in Process Models
07.09.09
Leave a Comment
You must be logged in to post a comment.
