Enabling theoretical model based techniques for simulating large scale networks

• Authors: Jennifer C. Hou;Hwangnam Kim
• Venue: Enabling theoretical model based techniques for simulating large scale networks
• Year: 2004

Modern data communication networks are extremely complex and do not lend well to theoretical analysis. It is common that network analysis can be rigorously made after leaving out subtle details that cannot be easily captured in the analysis. Instead, packet level, event driven simulation studies are usually resorted to better study the network performance. The major obstacle in packet mode simulation is, however, the vast number of packets that have to be simulated for producing accurate results. What seems to be a reasonable solution is to incorporate theoretical modeling into packet mode simulation. We develop a fluid model of describing the data transmission activities in IEEE 802.11-operated WLANs, and used it to explore whether or not fluid model-based simulation is effective in simulating WLANs. Fluid model based simulation is not well-suited for studying the network behavior under light and/or sporadic traffic, as it assumes a large number of flows in networks. To address the issue, we introduce the notion of network calculus based simulation; we characterize the interaction between TCP and AQM, determine necessary scheduling rules to regulate TCP traffic, and incorporate the rules into a simulation engine. Although both fluid model based and network calculus based simulation give encouraging results in terms of the execution time, they cannot provide packet level details, such as the instantaneous queue length and packet dropping probability, due to the use of larger simulation units. In order to provide the packet level dynamics, we propose mixed mode simulation. In mixed mode simulation, packet mode simulation co-exists with theoretical model-based simulation within one simulation framework. We also propose a new rescaling simulation methodology (RSM) to simulate IP networks with TCP and/or UDP traffic for the cases that the behaviors of all the flows should be inspected. The underlying idea of RSM based simulation is to reduce the computation by scaling down the network to one that can be simulated in a short time to produce sufficient results, and then to extrapolate the results for the original network with the results obtained from the smaller network.