Entropy-Based Dynamic Complexity Metrics for Service-Oriented Systems

The complexity analysis of service systems is mainly from the static aspect, and the dynamic complexity measurement has not attracted enough attention yet. In this paper, we attempt to provide some dynamic complexity metrics for service systems through measuring their execution behaviors. At first, two models, i.e., service unit vector and service pair matrix, are presented to represent the execution traces. Meanwhile, distance entropy is adapted in our work for analyzing execution traces. Based on the modeling representation of execution traces, the basic Shannon entropy and the adapted distance entropy are used to measure the dynamic complexity of a service system. In addition, the effectiveness of our proposed entropy-based metrics of dynamic complexity is validated by several examples of service systems and a real application. Particularly, the metric of combining the adapted distance entropy and service pair matrix has the strongest ability to depict the dynamic complexity.