Rate Control And Performance Optimization For Multimedia Service Over Internet

In the Internet era,multimedia service is changing people's life styles.It has been applied in many different areas like Internet Protocol Television(IPTV).video confer-ence.remote education,telemedicine,publishing,entertainment,electronic commerce and so on.Recent years have witnessed a rapid growing demand for multimedia service.The fact highlights the great market value and development prospect of multimedia ap-plications.But the dramatic increase of users' requirements poses immense challenges.As we known,smooth and high-quality multimedia services require sustained band-width.However,in reality both user behavors and channel status are time-varying.sto-castic and complicated,which make it hard to guarantee the Quality of Service(QoS).How to overcome these difficulities is the direction and focal points of this dissertation,and it is also a subject with realistic significance that should be solved.Besides,with the evolution of Internet technology,many new ideas have emerged,such as the decouple of control plane and data plane,the decouple of address and name,programmable routing etc,which have injected new vitality into the development of multimedia technology.How to combine the typical multimedia with these emerging network technologies is a major issue of immediate and long-term significance.The research contributes to im-proving users' QoS and bandwidth utilization,and thus is a new hotspot of developing future multimedia serivice.By using multimedia service as an entry point,this dissertation conducts a research on adaptive resource allocation and schduling based on the control and optimization theory.The main contributions of this dissertation are summerized as following:1)In wireless networks,user mobility and channel varation exhibit evident ran-dom.Adaptive media streaming is an effective way to solve this problem.In this dissertation,we carry out a research on adaptive rate control policy for on-demand scalable video service over wireless networks.We define Buffer Un-derflow Probability(BUP)for characterizing the mismatch between the video bitrate and the channel throughput.Accordingly,the adaptive Scalable Video Coding(SVC)transmission problem is formulated as the adaptive adjustment of video layers based on BUP.It jointly considers the buffer fullness and channel condition.Then the problem is formulated as a constrained problem.That is to optimize the attainable video quality,while keeping BUP below a desired level.In order to estimate BUP,we derive an analytical model based on the large devi-ation principles.Then,an online layer switching algorithm is proposed using this estimation model,which is capable of accommodating different channel qualities without any prior knowledge of the channel variations and of the video charac-teristics.We further introduce a perturbation-based layer sw itching approach for reducing the quality fluctuating issue caused by flicker effect,thus improving the viewer's Quality of Experience(QoE).A system prototype is implemented to evaluate the success of the proposed method.We also conduct simulations in multiuser scenarios with real video traces and the results demonstrate that the pro-posed algorithm is capable of balancing playback smoothness,fluency and video quality.2)By leveraging Software Defined Network(SDN).the intermediate network status,including instantaneous throughput,delay,packet loss,etc,can be observed di-rectly by the the controller.Such an architecture inspires us to utilize the collected network state by SDN controller to evaluate the network congestion dynamics over time.Consequently,based on the popular DASH technique,we carry out a research on network-aware adaptive multimedia service.The key problem is how to detect the network congestion as soon as possible and to probe the spare network capacity.The observed instantaneous throughput is highly variable over short time scales.To address this problem,we develop a rate adaptation algorithm based on a probabilistic indicator-Link Congestion Probability(LCP).Then we formulate a LCP-constrained optimization problem.The occurrences of the con-gestion are expected to be rare events,so we can use the large deviation princeple to evaluate LCP.However,during the evaluation,the prior knowledge of the dis-tribution of link traffic is necessary which is usually unavailable in reality.So we use Gaussian Mixture Model(GMM)to approximate this distribution.After-wards,we derive an online rate control algorithm.The experiment results on a prototype system demonstrate that with the controller's assist,our algorithm can improve the playback experience,while keeping a low playback stall rate and quality variation.3)With the aid of SDN controller,the forwarding path is programmable.It allows us to employ dynamic routing and multipath transmission in future multimedia applications.In this dissertation,after combining the typical rate control with dy-namic routing,we consider to develop a joint optimization policy in order to pro-vide a high-quality and smooth multimedia service.More specifically,if the link is congested at an intermediate node,the congestion can be detected by the con-troller.Afterwards,by searching the available paths,the controller selects a best path and updates forwarding tables of the switch.If no better path is available,video rate has to be decreased.In this dissertation,we use the Markov Decision Process(MDP)to model the optimization problem.To solve this problem,an on-line Q learning method is proposed,which avoids the prior knowledge of transfer probability.At last,we conduct some experiments on Mininet platform to verify our algorithm.Results show that the video quality and bandwidth utilization can be further improved in our system.