Quantitative evaluation of model consistency evolution in compositional service-oriented simulation using a connected hyper-digraph

Abstract

Appropriate maintenance technologies that facilitate model consistency in distributed simulation systems are relevant but generally unavailable. To resolve this problem, we analyze the main factors that cause model inconsistency. The analysis methods used for traditional distributed simulations are mostly empirical and qualitative, and disregard the dynamic characteristics of factor evolution in model operational running. Furthermore, distributed simulation applications (DSAs) are rapidly evolving in terms of large-scale, distributed, service-oriented, compositional, and dynamic features. Such developments present difficulty in the use of traditional analysis methods in DSAs, for the analysis of factorial effects on simulation models. To solve these problems, we construct a dynamic evolution mechanism of model consistency, called the connected model hyper-digraph (CMH). CMH is developed using formal methods that accurately specify the evolutional processes and activities of models (i.e., self-evolution, interoperability, compositionality, and authenticity). We also develop an algorithm of model consistency evolution (AMCE) based on CMH to quantitatively and dynamically evaluate influencing factors. Experimental results demonstrate that non-combination (33.7% on average) is the most influential factor, non-single-directed understanding (26.6%) is the second most influential, and non-double-directed understanding (5.0%) is the least influential. Unlike previous analysis methods, AMCE provides good feasibility and effectiveness. This research can serve as guidance for designers of consistency maintenance technologies toward achieving a high level of consistency in future DSAs.