Migration of processes from shared to dedicated systems : towards maintainable ad flexible processes

Abstract

Keeping multiple business processes deployed on one system poses severe threats to maintenance and flexibility of individual processes as well as the hosting system itself. Shared dependencies between processes limit the possibility to update the system for one process, because of the potential side effects on the other processes. The processes are bound to a system, even to a specific state of the system, and it is difficult to transfer them in a minimal way, so that only resources and dependencies of the respective process are considered. If documentation exists it is frequently outdated or complex to transform to an environment that is able to execute the process. Besides maintenance also preservation of the state of processes in a shared system consumes more resources than necessary. This makes it difficult to archive and share processes. The aim of this work is to provide a process migration framework (PMF) that frees processes from the chains imposed by the shared system where they are deployed. The first step of the PMF is to identify the process environment, so the resources that a process requires during execution, which is done by static and dynamic observation of the process and its environment. This process environment is stored in a model. The model can be adapted to upgrade software or migrate to different data types or other maintenance reasons. Next, a virtual system is created where the process environment is deployed. Finally it is verified that the target system is able to execute the process correctly. The result of the PMF is a documentation of the process environment as well as a virtual system that contains only resources relevant to the process. The advantages of this approach are that processes can be redeployed independently of any physical machine, processes are able to evolve independently of other processes, and process environments can be shared and archived in a sustainable way, including all dependencies that are required to execute the process. The evaluation of the PMF on different scenarios shows its ability to create models of the process environments at a level that is sufficient to recreate the process environment for scenarios that use workflows or scripts, which may invoke local tools and web services. The resulting target systems are able to execute the processes.