Cooperation of simulation model and data model for complex systems performance analysis

Abstract

Complex system is a system that can be analyzed into many components having relatively many relations among them, so that the behavior of each component depends on the behavior of others. It is intrinsically difficult to model due to numerosity, interactions, hierarchical organization. Modeling and simulation (M&S) is one of the fundamental methods of performance analysis of such complex systems. In other words, how well a modeler builds a model is a key point of a successful performance analysis. Before such a performance analysis, a model for prediction should be constructed. Therefore, two types of models have been studied: data modeling and simulation modeling. Data modeling is a method in which a model represents correlational relationships between one set of data and another. Conversely, simulation modeling is a more powerful method in which a model represents causal relationships between a set of controlled inputs and corresponding outputs. Since the complex system consists of many subsystems and components, it is difficult to model the whole system using only one modeling approach. For example, the data model can represent in detail through the actual data of the complex system, but it is difficult to analyze the system according to changes in the algorithm or model. Therefore, a methodology is required to classify and build a model of each subsystem/component by identifying its features. This dissertation identifies the limitations of each modeling approach and presents a cooperative model development process for performance analysis of complex systems. The cooperative method contains conceptual model design, model classification method, and model integration. The model classification method effectively reflects and maximizes the features compared earlier. Then, the classified models are modeled using proposed CoDEVS formalism that extends DEVS (Discrete Event Systems Specification). The formalism consists of simulation model, data model, and interface models that convert data between simulation and data model. Then, the models are integrated into the same simulation environment through the integration method. This dissertation also applies the proposed modeling method to develop a model of Hadoop using proposed CoDEVS formalism and ANN (Artificial Neural Network). To demonstrate the validity of the case study, it presents experiments to show the advantages of the cooperative modeling and shows the possibility of a proposed modeling approach that differs from the existing approaches. We then showed how the proposed modeling approach can be applied in the real system, and how it can create synergy through the cooperation between two models for performance analysis of complex systems.