Design And Implementation Of A Self-adaptive Workflow Management Framework Based On Bpel

With the social and economic development, enterprise level applications become more and more complicated, and the internal enterprise business process management seems especially important. To deal with these problems, there appear a large number of workflow management systems. But current static workflow systems gradually appear shortage in front of demands which changes rapidly. What we need is an open, dynamic, personalized computing environment, which we can quickly deal with frequent process changes within.BPEL is the industry standard description language of workflow, and it’s supported among multiple engines. Then we propose a variable process description language based on BPEL and a self-adaptive management framework. All of these are designed on the principle of software reusing and portability of system application. At the end of this paper, we implement the framework combing with current JBPM-BPEL system and take some tests to display the correction and stability. Any workflow engine (based on BPEL) which implements the interfaces of framework will support the variable process model seamlessly. This paper mainly focuses on three aspects:1. Put forward a variable process describion languageWe extend BPEL by adding variable points. And the variable process model described by this extented language is the parent of all specific models.While executing the variable process model will be translated into a standard BPEL model. That is to say, we adopt the concept of process family, and divide models into abstract and specific ones in terms of variable points. Abstract process model describes the unknown situations, and specific model is an abstract model whose varable points are filled.2. The design of self-adaptive workflow management frameworkThe framework covers the following main parts:①The deployment module which changes variable process models to standard BPEL models, maintains the relationship between variable points and filling informations.②The monitoring module obtains the context information of the engine, and triggers the fulfillment of variable points. During handling a processs, it truns internal execution state to specific events (state normal, abnormal, meeting variable point), and then throw it out, which is catched by a registed event listener.③The model management module. Its responsibility is to manage running process instances and process models. Such as suspending, recovering and destroying of instances.④The migration module whose main task is to shift a running instance from its dependent model to another. When an instance is under execution, it firstly searches all running instances of a specific process definition and stops them. Secondly, each of them should be judged whether it can be migrated or not. On migration examples we calculate its target state, and then move the running data from the old instance to a new instance. Otherwise it resumes the instance, executing or disrupting in accordance with the old model.⑤The interactive module gives out filled slices for variable points. We use AI planner in this implementation, which plans BPEL fragment based on context information. This module works as a black box, only concerning about inputs, so its implementation is loosely coupled with the other modules of the self-adaptive management framework.⑥Log management module. This module records runtime state information of system, including the moment of variation points being replaced, and the filling of variable points, the start and end time of instance migration, as well as the execution information of migrating instances and system runtime exceptions.3. The implementation of the self-adaptive workflow management frameworkThe implementation of framework is the final part of our researches. It consists of the following three parts:①Propose interfaces which act as a foundation of other workflow systems that realize this framework②Use JBPM-BPEL as the workflow system, and implement this framework.③Modify JBPM-BPEL engine to implement the interaction between workflow system and self-adaptive management layer.