Traffic characteristics of aggregated module downloads for mobile terminal reconfiguration
In recent years, there has been an increasing proliferation of reasons for software/firmware in desktop computers to be upgraded over the Internet, and as computational devices continue to become ever more advanced and compact, it is anticipated that the same will be the case for mobile terminals in the near future. In this paper, we consider some important characteristics of such upgrade traffic. Initially, we study file size distributions for various file types on a range of servers/systems, and conclude that the Pareto distribution, as is often used to represent file sizes, is inappropriate. In fact, in order to achieve a better fit to the bulk of the distribution while still maintaining an infinite variance, file sizes are usually far better represented by the Pareto type 2 (Lomax) distribution. Leading on from this, we simulate multiple concurrent files being downloaded over a bottleneck link, taken from particular file-size distributions with a focus on mobile terminal reconfiguration. In doing so, we investigate the degree of burstiness/self-similarity and correlation structure in the resulting traffic. Understanding these phenomena is particularly important because of their tendencies to increase packet losses and queuing delays through compound fluctuations in link loads.
