Affiliations: Department of Computer Science and Engineering, University of California, Riverside, CA 92521, USA E-mail: {jobin,michalis,tripathi}@cs.ucr.edu
Abstract: The goal of this paper is to develop a systematic approach to conducting simulations in wireless cellular networks. Performance analysis in such networks is a non-trivial task for the following reasons. First, there are a lot of parameters whose combined effect needs to be understood. Second, there are no benchmarks or widely accepted realistic values. As a result, different studies use different scenarios, and this makes the results of these studies incomparable. How can we make network scenarios and results easier to interpret and compare? We start with the observation that absolute parameter values are meaningless. Instead, we propose the use of relative parameters. These parameters combine multiple parameters into fewer parameters; e.g., user speed and diameter of cell are combined into the cell latency parameter. Thus, we reduce the total parameter space, but more importantly, we capture the combined effect of parameters that affect the performance. We show how using relative units of time and space and relative parameters such as cell latency and average number of cells visited can provide better insight than absolute parameters such as session duration or absolute units such as seconds, miles, etc. In addition, we propose the use of steady state utilization, a novel metric that captures the inherent capacity of a mobile network. We show that steady state utilization can become the reference point with which we can conduct meaningful simulations and interpret results more effectively. Finally, we describe a case study to illustrate our methodology. We evaluate the effect of bandwidth reservations on network performance. Our results show how reservations can actually hurt network performance.
