Since timely completion of business process in system-oriented workflow environment is important for Quality of Services(QOS), the Workflow Management System needs to provide the facility to define and manage the deadline of workflow processes.
A workflow consists of a set of activities that are individual process steps and agents executing these activities. Many escalations may occur if some agents are highly loaded and require much time to complete the activity. To prevent this situation and improve performance, we can consider the deadline of each activity and invoke, if necessary, early escalation to minimize escalation cost and allow more system resources to other processes. Some mechanisms for managing activity deadlines during process execution have been proposed. They commonly assume the deadline of each activity is pre-defined by business analysts. Typically, business analysts, however, only focus on the entire process deadline, and may not know about the information of the deadline of each activity.
In this paper, we propose a new slacktime allocation method that allocates slacktime proportional to the sojourn time of each activity. We estimate the mean sojurn time of each activity and analyze workflow control structures such as And, Or, Iteration to estimate the mean sojourn time in workflow. We use queueing theory and the probability density function of each activity's sojourn time to obtain the expected value of the additional slacktime. We only consider the case that each branch is composed of just one server, but we also present a mechanism to transform multiple servers to one server approximately. In our experiments, we validate the estimates of sojourn times and show our method is more efficient than other slacktime assignment methods
