(A) joint control of multiplexing and traffic smoothing for multiple VBR MPEG videos

Abstract

In the near future, it is expected that VBR MPEG video service is one of the most potential applications in ATM networks because of the high bandwidth and the flexible bandwidth allocation of ATM technologies. Besides, the transmission technique of VBR MPEG video will form the basis of the advent multimedia services over ATM network. In order to facilitate the introduction of VBR MPEG services over ATM network, a joint control algorithm of multiplexing and traffic smoothing is investigated in this dissertation for both live and prerecorded VBR MPEG video streams. First, we investigate the issues related to the source traffic smoothing for a live VBR MPEG video over ATM network. We present a real system model for the traffic smoothing of live VBR MPEG video applications. Based on this system, we describe the constraints imposed not only by the delay bound in the sender buffer, but also by the sender/receiver buffer sizes in the viewpoint of practical parameters. These constraints are formally defined as the upper and lower bounds for the decision of transmission rate, respectively. Then, by using these constraints, an adaptive source traffic smoothing algorithm is designed in a dynamic fashion, while avoiding the buffer overflow and underflow. Through simulation results, it is shown that the proposed algorithm result in small loss rates, in the case that overall end systems has extremely small buffers. Under the various values of buffer sizes and delay bound, the proposed algorithm reduces the temporal variability as well as the peak rate of the given live video, compared to the sliding window scheme. When the simple queueing models based on the large deviation theory are applied, the proposed algorithm significantly reduces the renegotiation failure probability as well as the effective bandwidth of the given live video, compared to the sliding window scheme. Furthermore, it is noted that by developing more exact coded-data size estimation method, the performan...