Research On Routing Mechanisms In Network Simulation

The last decade has seen the rapid growth of Internet and the great impact it has made to the social life. In order for better development of Internet, it's quite important to deeply understand the behaviors of Internet. Due to the complexity of networks, there are more and more difficulties in modeling Internet. Simulation plays a vital role in attempting to characterize both the behavior of the current Internet and the possible effects of proposed changes to its operation. The use of simulation is becoming increasingly prevalent in the networking research community.Routing is the most important among many factors that affect simulation scale and efficiency. Routing mechanisms in network simulation can greatly affect the fidelity, scale and efficiency of the simulation. Thus, how to simulate the routing behaviors is a key problem in network simulators, and an important way to improve the simulation scale, efficiency and fidelity.Based on the background mentioned above, this dissertation places emphasis on improving the simulation efficiency and fidelity, and the following aspects are mainly included.Firstly, the abstract model of network simulation is given. Then the dissertation presents the abstract models of routing mechanisms for network simulation. And also the design principles and abstract methods for the routing mechanism in network simulation.Local static routing mechanisms are the foundation of the routing mechanisms in network simulation. Till now all the current routing mechanisms have their own disadvantages. After analyzing some of the current routing mechanisms, the dissertation presents an efficient routing computation, storage and lookup approach, named MTree_Nix. MTree_Nix integrates the advantages of some well-known routing mechanisms, maintains variable number of spanning trees as the base routing table for most of the shortest-path routing states, and for those that can not be covered by any of the spanning trees, MTree_Nix uses Nix-Vector routing to compute them on demand. By analyzing the storage space requirement and CPU time for routing states lookup, the dissertation gets the best trade-off between them, and the constraint condition to achieve this. Experimental results show that MTree_Nix routing indeed can achieve a better trade-off between the memory usage and simulation time, and much more appropriate for the simulations of complex applications on large-scale topologies.Remote static routing computation and lookup is one of the key factors of the efficiency of parallel network simulation. After analyzing the routing mechanism that is the most widely used in most network simulators, the dissertation presents a new method which computes remote routing based on the border routers, and finds the appropriate route by comparing the length of routing. Experimental results show that this mechanism effectively reduces the size of remote routing table, memory usage for the initialization of the simulator, and the runtime of the simulation. And it greatly improves the efficiency of network simulation.Dynamic routing mechanism is a key problem when it needs to consider the effects of topology changes in the simulated application. This dissertation analyzes the problems of dynamic routing according to the dynamic routing model, which are the awareness time of topology changes at different nodes, and the dynamic routing lookup algorithm. Then a method to determine the awareness time of different simulated nodes is presented. Based on the static routing table in MTree_Nix mechanism and the queue of topology changes, with the comparation of current time and the awareness time of the node, the dissertation presents dynamic MTree_Nix routing mechanism to look up the routing states, so that the dynamic routing problem is solved.Worm propagation is one of the most serious problems in network security. Simulation is an efficient method to study the behavior of Internet worm propagation. Based on the routing mechanisms mentioned above, the dissertation presents a packet-level Internet worm simulator, called CCDRWS (Congestion Control and Dynamic Routing based Worm Simulator), which introduces the congestion control model into the simulation environment to decide node and link failures based on the traffic load. CCDRWS also uses dynamic routing mechanism to update routing states, and keeps high simulation efficiency.