Queueing networks approach for bandwidth estimation of smoothed VBR video streams

The growing importance of multimedia applications, which usually implies the transmission of variable bit rate (VBR) video streams, needs an accurate analysis of bandwidth requirements according to specified quality of service (QoS) guarantees. In this paper an original algorithm is proposed for statistical aggregate bandwidth estimation of smoothed VBR video streams, respecting a specified loss probability (the QoS parameter taken into account). This algorithm is based on concepts of multiclass networks of queues whose service centers are represented by the bandwidth levels assumed by smoothed films. The mean service times and the transition probabilities between service centers are directly derived from real video traces. The analytical results have been compared with simulation results and with the results obtained by another analytical method already present in the literature, i.e., the Chernoff bound, discussing the effectiveness of the proposed method.