QoS provisioning in multimedia streaming

Multimedia consists of voice, video, and data. Sample applications include video conferencing, video on demand, distance learning, distributed games, and movies on demand. Providing Quality of Service (QoS) for multimedia streaming has been a difficult and challenging problem. When multimedia traffic is transported over a network, video traffic, though usually compressed/encoded for bandwidth reduction, still consumes most of the bandwidth. In addition, compressed video streams typically exhibit highly variable bit rates as well as long range dependence properties, thus exacerbating the challenge in meeting the stringent QoS requirements of multimedia streaming with high network utilization. Dynamic bandwidth allocation in which video traffic prediction can play an important role is thus needed. Prediction of the variation of the I frame size using Least Mean Square (LMS) is first proposed. Owing to a smoother sequence, better prediction has been achieved as compared to the composite MPEG video traffic prediction scheme. One problem with this LMS algorithm is its slow convergence. In Variable Bit Rate (VBR) videos characterized by frequent scene changes, the LMS algorithm may result in an extended period of intractability, and thus may experience excessive cell loss during scene changes. A fast convergent non-linear predictor called Variable Step-size Algorithm (VSA) is subsequently proposed to overcome this drawback. The VSA algorithm not only incurs small prediction errors but more importantly achieves fast convergence. It tracks scene changes better than LMS. Bandwidth is then assigned based on the predicted I frame size which is usually the largest in a Group of Picture (GOP). Hence, the Cell Loss Ratio (CLR) can be kept small. By reserving bandwidth at least equal to the predicted one, only prediction errors need to be buffered. Since the prediction error was demonstrated to resemble white noise or exhibits at most short term memory, smaller buffers, less delay, and higher bandwidth utilization can be achieved. In order to further improve network bandwidth utilization, a QoS guaranteed on-line bandwidth allocation is proposed. This method allocates the bandwidth based on the predicted GOP and required QoS. Simulations and analytical results demonstrate that this scheme provides guaranteed delay and achieves higher bandwidth utilization. Network traffic is generally accepted to be self similar. Aggregating self similar traffic can actually intensify rather than diminish burstiness. Thus, traffic prediction plays an important role in network management. Least Mean Kurtosis (LMK), which uses the negated kurtosis of the error signal as the cost function, is proposed to predict the self similar traffic. Simulation results show that the prediction performance is improved greatly as compared to the LMS algorithm. Thus, it can be used to effectively predict the real time network traffic. The Differentiated Service (DiffServ) model is a less complex and more scalable solution for providing QoS to IP as compared to the Integrated Service (IntServ) model. We propose to transport MPEG frames through various service classes of DiffServ according to the MPEG video characteristics. Performance analysis and simulation results show that our proposed approach can not only guarantee QoS but can also achieve high bandwidth utilization. As the end video quality is determined not only by the network QoS but also by the encoded video quality, we consider video quality from these two aspects and further propose to transport spatial scalable encoded videos over DiffServ. Performance analysis and simulation results show that this can provision QoS guarantees. The dropping policy we propose at the egress router can reduce the traffic load as well as the risk of congestion in other domains