Extended GSPN modeling and reduction algorithms for rapid performance analysis of service composition system

A service composition system can be constructed via integration and collaboration of many component services. The system performance needs to be obtained rapidly for the purpose of dynamic and adaptive management. An EGSPN (Extended Generalized Stochastic Petri Net) model is presented to reflect the service composition system with complex timing constraints in a compact and comprehensible manner. The various categories of timing constraints are associated to place, transition and arc of EGSPN respectively. Furthermore, in order to accelerate the calculation of system performance and to avoid the explosion of state space, a set of model reduction rules are presented. The corresponding reduction algorithms are designed to achieve automatic model reduction. The experiments based on multiform models indicate that a large model can be reduced within satisfying period. The algorithms can help to make rapid performance analysis of service composition system.