Task net: Transactional workflow model based on colored Petri net

Task net: Transactional workflow model based on colored Petri net

0.00 Avg rating0 Votes
Article ID: iaor20022828
Country: Netherlands
Volume: 136
Issue: 2
Start Page Number: 383
End Page Number: 402
Publication Date: Jan 2002
Journal: European Journal of Operational Research
Authors: , ,
Keywords: petri nets
Abstract:

Workflow management systems are increasingly identified as a tool that can not only integrate existing computer applications but also dynamically change existing business processes. Unlike database transactions that have a mechanism to guarantee consistency against failures or concurrent execution, there is no known mechanism that guarantees safety of workflow transactions. Therefore, capability to forecast the effects of workflow transactions is very important. That is, a model is required which allows to determine whether a workflow transaction can terminate, whether it is free from dead lock, whether it can recover from error or crash, etc. In this paper, we propose Task net which is a transactional workflow model based on colored Petri net. First, we introduce a workflow specification language that can express task state dependency relationships such as forward, backward, and mutual exclusion among intra-task states (states of the same task) and inter-task states (states of different tasks). The proposed workflow language will enable the users to express both the transaction- and application-oriented requirements of complex business rules. We, then, describe how a basic workflow specification and its task state dependencies are translated to a Task net which will allow simulation of the given workflow. Before executing a Task net, tests for deadlocks and redundant task state dependencies are performed to prevent futile efforts. Task nets are expected to provide a theoretically sound basis for current workflow systems to expand their transactional workflow supports.

Reviews

Required fields are marked *. Your email address will not be published.