Quantitative assessment of workflow performance through PH reduction

Workflows are logical abstraction of processes widely adopted in several contexts such as economy and management sciences (business processes), service engineering (service oriented architecture, Web services, BPEL), software engineering (component based systems, UML, flowcharts) distributed computing (Grid, Cloud, Mapreduce). Design and operation of workflows are critical stages in which problems and issues not manifested by the single block arise from compositions. To deal with such issues, proper techniques and tools should be implemented as support for workflow designers and operators. This paper proposes a solution for the evaluation of workflow performance starting from the components’ ones. Based on the stochastic characterization of the workflow tasks, phase type distributions and stochastic workflow reduction rules, the proposed approach allows to overcome the limits of existing solutions, considering general response time distributions while providing parametric analysis on customer usage profiles and design alternatives. To demonstrate the effectiveness of the proposed solution an example taken from literature is evaluated.