On packet loss modeling: An empirical assessment

Accurately modeling the packet loss behavior of network components is a longstanding problem due to the impact of loss on quality of service (QoS) management and optimization strategies for real-time multimedia applications. Although a few models have been proposed in the literature with the purpose of estimating the statistical behavior of packet loss processes, none of them correctly reflect the property of a network to form both bursts of consecutive losses and aggregates of single (and/or small bursts of) losses occurring at short distance from one another. In this paper we propose a new model that incorporates the characteristics of burstiness and aggregateness observable in a sequence of packet loss events. We also provide a comprehensive comparison of it and the current state-of-the- art models with traffic traces routed by a bottleneck node. Our results demonstrate our proposed model's ability to outperform the existing models, due to its capacity to capture this double essence of burstiness and aggregateness.