Study on using design patterns to implement a simulation system for WiMAX network

Abstract

A network simulator always is an important tool to observe and evaluate the study concept for wireless networks. Considering the restriction of the limited budget, the non-commercial open-source simulation software often becomes the top choice of academia. Since the natures of academic study are innovation and excellence, a simulator frequently encounters that the existing modular functions are inadequate and need to be appended or modified. If a selected simulator inherently has poor architecture, the maintenance, recondition, and expansion of functions will become more difficult and more time-consuming in the future, while it is difficult to understand and reuse by the successors. Therefore, how to select a most suitable simulator is an important issue. In order to make a simulator have flexible architecture and believable results, design patterns are proposed as the norms to design system architecture. To realize this idea, we surveyed six most used simulators, i.e., J-Sim, NS-2, NS-3, OMNet++, OPNET, and QualNet, to ponder their system architectures and design concepts from the source codes and the related literatures of the modular function expansion. We propose a network simulator architecture, named as CCGns, which is a discrete-event virtual network simulator and follows the IEEE 802.16-2009 standard. CCGns obeys the object-oriented design principles and is coded by the Java language. CCGns comprises eight packages for physical layer, medium access control layer, network layer, devices, topologies, events, scheduler, and reports, respectively. The main contribution includes three aspects which propose a scalable MAC messages management and the corresponding architecture, an applicable for multi-hop relay network topology architecture, and a two-stage minimum variance bandwidth allocation algorithm. To the best of our knowledge, this article is not the first one to apply the design patterns for the simulator architectures of wireless network, but we use more design patterns and types than the others and also provide the unified modeling language figures to explain the system architectures. We particularly focus on how the management procedure of control messages influences the time-related performance evaluation metrics, e.g., how the amount and processing time of different control messages affect throughput, packet delay, and packet drop ratio. By using mathematic calculation to verify the simulation results, the proposed system architecture has been proven to possess excellent fidelity.