Reducing the number of migrated instances during business process change: A graph rewriting approach.

[Display omitted] • The goal of this contribution is to reduce the number of migrated instances during business process change without affecting the number of instances that satisfy the new requirements. To achive this goal: • We propose the use of the Petri Nets (PNs) formalism and Graph Rewriting Rules (GRRs) to model BPs and their changes, in which a PN is used to model the current BP, and GRRs are used to model the BP change and the instance migration. • Our lazy strategy considers two consistency criteria to ignore the migration of instances that satisfy the new requirements. • We implement our approach in the framework, Flexible (Bouhamed et al., 2021). • We illustrate our approach by means of a running example modeling a real-world scenario where two BP changes are introduced. In addition, we compare our approach with a brute-force approach not only in this example but using five other BPs proposed in the literature as well. Business processes (BPs) are subject to frequent and unpredictable changes caused by new laws or modifications in the business environment, including user requirements. Many running instances are migrated to new BPs during a BP change, even when these instances satisfy the new requirements. We propose a migration strategy to ignore the migration of these instances. We specify the BP using Petri Nets (PNs), and the change including the instance migration using Graph Rewriting Rules (GRRs). We propose two consistency criteria based on the instance execution history and its possible future executions to identify the instances supporting the change. The first criterion ignores the migration of the instances satisfying the new requirements. The second criterion postpones the migration of instances that support the migration until they definitely cannot satisfy the new requirements, where some instances might avoid the states where the migration is mandatory. The proposed approach is implemented through the framework called Flexible and illustrated using two cases of BP changes. Simulation results shows that a large number of instances can be ignored during the migration, 86.34% and 42.35% respectively. The approach is evaluated using five real-life scenarios, where a positive correlation with respect to the complexity is observed. [ABSTRACT FROM AUTHOR]