Research On Temporal Coordination And Failure Recovery Of Artifact-centric Business Processes

With the growing competition in the international market, business processes have be-come a necessity for modern organizations to stay competitive. A business process assembles a collection of tasks or activities to achieve a business objective. How to ensure the reliability and success of a process execution is very crucial for business process management. In the service-oriented environment, long running business process will inevitably encounter various types of errors, such as, services are not in their available time interval, service delay or can-celation by users. Business processes fail to accomplish when temporal inconsistency occurs between participants or failures are encountered. Therefore, temporal consistency management and failure recovery are intensively under investigation.There have many research efforts on managing temporal consistency in business collab-oration. However, most of these works only focus on available temporal information of ser-vices, ignoring the influence of data on temporal consistency; or specify too strict temporal policies(E.g., all previous services are required to finish before the last available time point of a service) which do not apply to many real-life scenarios. However, data has a great influence on temporal consistency, for example, a service provider always sets up different available time span for a service, or discover available temporal information of entities instead of related ser-vices, such as, credit record of a customer. Therefore, it needs to consider the impact of both service and data on temporal consistency management.When temporal inconsistency or other failure occurs during the execution of a business process, it needs to be recovered to ensure execution correctness of the process. The recovery can be forward or backward. Unfortunately, backward recovery of a business process will inevitably lead to significant financial loss. In real-life commercial activities, the participants of a collaborative activity always hope to make the process forward recovered as far as possible through technique methods or proper financial compensation to achieve business objective. At the same time, they also hope to get some compensation from the fault party when the process has to roll back. Note that process instances may overlap and interweave with each other which significantly complicates the failure recovery issue. To forward recover a failed process as far as possible meanwhile decrease the failure impact on other concurrent processes, a new recovery mechanism is required to combine the classical forward recovery techniques with financial compensation. Complicate cardinality relationship needs to be considered to avoid unnecessary recovery.To solve above problems, it needs to give equal consideration on modeling both data and control flow of activities and shift from activity-centric models towards data-centric models, in which artifact-centric business process model is one of the typical representatives. In this thesis, we mainly concentrate on temporal consistency management and failure recovery, based on artifact-centric business process model. The main contributions are as follows.First, this thesis proposes a model named Timed Artifact-centric Net (TiArtiNet), consid-ering temporal information for both service and artifact. Based on TiArtiNet, the corresponding algorithms are developed for checking temporal consistency at both design time and runtime. By analyzing temporal consistency for all reasonable branches combination, we improve the accuracy of temporal consistency verification and the success rate of temporal coordination. Experimental results show that the proposed approach can address the challenging issues of temporal consistency in artifact-centric business process collaboration correctly and efficiently.Second, this thesis proposes a context-aware business process transactional model, which supports the definition of conditional alternative services and financial compensation rules be-tween partners. Based on the model, a forward recovery mechanism, which combine financial compensation and conditional substitution, is proposed to handle different kinds of exceptions with different recovery strategies to make the process forward as far as possible.Finally, to address the failure recovery issue of the fabric business processes in relevant business scenarios, this thesis proposes a best-effort failure recovery approach MaxInsTx that protects maximal instances involved in failures from failure impact. The approach includes: (1) a transactional artifact-centric business process model with cardinality and correlations between artifact classes explicitly specified; and cardinality types of process dependencies dis-tinguished; and (2) an approach to resolve the impact of the failed process on its dependent process instances. To avoid unnecessary recoveries, overlaps are used to determine if a de-pendent instance is really affected. Three categories of overlap degrees are defined, including no overlap, partial overlap and complete overlap, and recovery strategy for each category is provided. An algorithm for recovering affected instances is given and its effectiveness and feasibility is verified by experiments. MaxInsTx is the first work to resolve the impact of fail-ure recovery of one instance on its dependent instances meanwhile avoid unnecessary recovery using correlations and overlaps among instances.