A routing algorithm for distributed simulation systems

Keywords: Modeling & Simulation, DEVS, Network Management, Ecology, Internet Network systems must communicate with one another using a variety of algorithms and technologies. Many systems supporting interconnectivity are required to exhibit essential traits such as adaptability, scalability, and reliability (survivability). At the same time, these networked systems are expected to offer new and more sophisticated services in the face of increasing system heterogeneity. Consequently, to cope with the management of such networks in the presence of ever increasing complexity, various decentralized and centralized approaches are being used to address the needs of private and public organizations. In this thesis, in order to examine some problems of distributed systems such as complexity and scalab resolution levels, a modeling and simulati methodologies and tools are used. Scalabil lity, search for design alternatives and propose various on study is performed in which modeling and simulation ty issue has become very crucial concept for modeling and simulation of the Internet. Simulators generating static topology such as COMNET, NS2, and OPNET are ideal platforms for studying small networks, but incapable of modeling and testing large-scale and dynamic structure networks. Furthermore, due to lack of hierarchy and abstraction in their structure, it is difficult to create and manage large model families. In this study, a network simulator is developed to bring solutions to above problems by using DEVS modeling and simulation methodology. To develop and study dynamic and adaptive swarm-based routing protocols of biological, we have devised a DEVS (Discrete Event System Specification) network model. The nodes and links are characterized as the elementary network components. These models and networks are implemented in the DEVSJAVA modeling and simulation environment which is an implementation of the DEVS framework. The developed network environment is capable of representing behavior of different routing algorithms (e.g. shortest-path, distance vector and other swarm algorithms). For the purpose of modeling of a distributed networked system, we define a network in terms of its components (e.g. nodes and links) and their hierarchical structure. The network is then modeled as DEVS atomic and coupled models. To do this we closely examine biological aspects of honeybee colony and their mappings into DEVS models. For example, the routing policy of the network which is embedded in every node is similar to a honeybee having its own capability to search for food and communicate with other honeybees. Using the DEVS hierarchical model composition concept, we develop simulation models of networks with varying topologies and scales. For example, we will use clusters to study its impact on reducing communication and increasing performance. The explicit and hidden behaviors of these networks are observed under various experimental configurations - e.g., nodes and links are assigned different capacities. In this thesis, performed operations can be summarized into four steps: i- Modeling and simulation of distributed systems together with nodes and links components using DEVS. ii- Developing a rule-based swarm intelligence routing algorithm by inspiring from honeybee scout- recruit system. iii- Modeling a state-of-the-art routing algorithm in order to depict advantages of developed algorithm. iv- Creating large-scale networks models having from tens to several thousands of components and observing the behavior of developed algorithm. Xlll