QoS Analysis of Self-Similar Multimedia Traffic with Variable Packet Size in Wireless Networks

Accurate modeling of multimedia traffic is important for optimizing resource allocation such that Quality of Service (QoS) can be most satisfied with constrained resource in wireless networks. Multimedia traffic is demonstrated to exhibit self- similar nature. We take account of this kind of characteristic and propose a comprehensive analytical model subject to variable capacity in wireless networks from the viewpoint of queueing theory. Different from existing work on QoS analysis of self-similar traffic, we focus on practical multimedia traffic where packet size is variable. In the proposed analytical model, the self-similar behavior of multimedia traffic is simulated by the fractional Brownian motion (fBm). Furthermore, packet size is assumed to follow the heavy-tailed distribution (e.g. Pareto distribution). Finally, the closed-form expressions for QoS metrics (i.e. the tail distribution of queueing delay and packet loss probability) for self-similar traffic with variable packet size are derived. Our performance evaluation shows that the results from our analysis are consistent with those from simulations. Also, the effect of packet size on QoS metrics is carefully explored.