Time Performance Analysis For Instances Of Generalized Well-formed Workflow

Nowadays, enterprises face the fierce market competition and frequently-changing demand for services, thus its internal architecture and business processes must be adjusted regularly to meet the challenge. Workflow technology, because of its high efficiency, automatic task scheduling model and ability to adapt to business process reengineering, has been receiving much attention. In implemented workflow management systems, completion time of running instances of business process is a key issue that receives much concern from both enterprises and its clients because it has a direct impact on both relevant work arrangements of enterprises and customer service quality.This thesis focused on several key problems in predicting completion time of business process instances.In this paper, by analyzing shortage of existing well-formed workflow models, generalized well-formed workflow model (GWFWM) was put forward to extend the ability of workflow model to describe business process; also, an algorithm to build workflow nets in conformity with GWFWM was proposed. In order to facilitate automatic calculation of time performance for generalized well-formed workflow, method to compute routing probability of instance subgraphs in workflow nets was proposed, and rules to identify basic workflow patterns were given. Lastly, an algorithm to analyze time performance for generalized well-formed workflow was put forward.To analyze time performance for running workflow instances, first of all, states of transitions were classified, and different methods to compute time performance for transitions in different states were proposed. Specially, for executing transitions, active transition performance equivalent models were proposed to compute their time performance; furthermore, performance models for workflow patterns including active transitions were deduced.For workflow nets in environments subject to changes, the concept of reachable subnets was proposed, and an performance analysis algorithm based on technology to decompose subnets was given, in which generalized well-formed reachable subnets were transformed into more simple instance subgraphs, thus existing reducing approaches can be used to compute time performance for it. Also, computing complexity of the algorithm was analyzed, and optimizing approach was proposed.For workflow nets relatively stable in structure and resources, concepts of potential instance subgraphs set, inflexion set of instance subgraph, inflexion set of workflow net, and inflexion set of workflow instance were proposed, and the Corresponding relationship between workflow instance and its inflexion set was demonstrated, thus an more easy approach to find potential instance subgraph set for workflow instance was deduced. By avoid stages to build and decompose reachable subnets, this approach can reduce time spent in performance analysis for workflow instances.Waiting time of workflow instances at activities was important part of its completion time, which varied dramatically dependent on system load. In this paper, a method to estimate waiting time of workflow instance at every activity according to current load of whole system was proposed. This approach was based on Monte Carlo simulation. Waiting time in tandem queuing systems can be estimate by a simulation program designed for tandem queuing systems with multiple service stations. Repeated simulation under same environment settings achieved consistent results, indicating the availability of the simulation method.Based on the research on several key aspects of time performance analysis for workflow instances, this paper try to build a sound performance analysis framework to improve quality of service and work arrangement for enterprises.