Pieter Hens

Process fragmentation, distribution and execution using an event-based interaction scheme

HighlightsA deployment and execution architecture is presented that can automatically partition and execute a given process model, according to a predefined process distribution.The resulting partitions achieve a natural autonomous task decoupling.Collaboration between the partitions is event based.The coordination overhead of the distributed process execution architecture is minimal.The distributed execution still allows process change. The combination of service oriented architectures and business processes creates an enactment environment in which processes can be deployed and executed automatically. From a managerial and technical point of view, the interpretation, control and execution of a process flow happen very often at one point in the organizational and IT structure. This creates an inflexible environment in which control over and visibility of cross-departmental processes cannot be distributed across these organizational entities. Although the process model may need to be designed as a whole (to have an end-to-end definition), the actual execution of the process may need to be distributed across all participating partners. There are several ways to achieve this distribution. In this paper, we look at an event-based process deployment and execution infrastructure in which a process model can be automatically partitioned and distributed over different enactment entities, provided some given distribution definition. We compare the performance and flexibility of the proposed technique with other approaches and discuss the potential advantages and drawbacks of the event-based distribution.

Verification of Change in a Fragmented Event-Based Process Coordination Environment

To allow the distribution of control and visibility of cross-organizational process models and to increase availability and performance of the processes, a process model can be fragmented into logically different parts and distributed in the enterprise architecture. Fragmentation algorithms and execution environments that connect the fragmented process model parts together, recreating the original process execution semantics, have been proposed in earlier works. However, a critical challenge that is left open is the ability to independently change the control structure of a process fragment. This is not trivial as changing the control structure of a specific fragment could break the global process execution. The global process overview is also not available anymore because it is fragmented in the enactment environment and many independent control structure changes could have already been done. In this paper, we describe an approach based on state reconstruction, which enables checking the admissibility of changes made to the control structure of a process fragment in a fragmented event-based process enactment environment.

Decentralized event-based orchestration

Today, in the state of the art process engine solutions, process models are executed by a central orchestrator (i.e. one per process). There are however a lot of drawbacks in using a central coordinator, including a single point of failure and performance degradation. Decentralization algorithms that distribute the workload of the central orchestrator exist, but they still suffer from a tight coupling and therefore decreased scalability. In this paper, we aim to investigate the benefits of using an event driven architecture to support the communication in a decentralized orchestration. This accomplishes space and time decoupling of the process coordinators and hereby creates autonomous fine grained self-serving process engines. Benefits include an increased scalability and availability of the global process flow.

Measuring the Impact of Suspension on the Process Enactment Environment during Process Evolution

Current workflow management systems implement the ability to automatically execute predefined process models. However, processes change over time and therefore a redeployment process has to be implemented to propagate changes into the running process enactment environment. One of the necessary steps in change propagation is to suspend the current process execution. This suspension does however decrease availability of the workflow management system, increases downtime and implicitly also decreases scalability. In this paper we provide a quantification of the impact of suspension on the runtime process enactment environment and experimentally evaluate this impact, hereby providing a better insight in suspension impact. Furthermore two suspension techniques are compared and a discussion is provided in which situations, which suspension technique is beneficial.

Distributed Event-Based Process Execution - Assessing Feasibility and Flexibility

Processes modeling and execution (with a process engine) are getting more and more incorporated in todays business environments. This movement puts a lot of stress on classical process engines which have to coordinate many process instances simultaneously. Performance degrades quickly as the number of process instances increases, and a single point of failure is introduced by using a central process execution engine. In this paper, we address these challenges by providing a non-intrusive approach to distribute a process flow and have the flow executed by multiple, smaller process engines. We pay special attention to flexibility of the eventual distributed execution, since process change is costly in a distributed environment. We demonstrate the feasibility of our approach by providing an implementation of the transformation and execution architecture, and demonstrate the lower cost of process change that is achieved when using a flexible process runtime architecture.