On Broadband Access And Congestion Control Of Streaming Media Services

Triple-Network Convergence is the trend of the information infrastructure evolution. Stream media, as a kind of rapidly arisen application, is considered as a catalyst in achieving Triple-Network Convergence. Network convergence and stream media services development necessarily demands better performance and QoS supplied by the networks. Therefore, it is of great theoretical and practical significance to research the problems in the process of network convergence and streaming media development.In this thesis, we focus on the issues of broadband access and congestion control technologies of streaming media services on the background of network convergence and streaming media development. The main works and innovations are as follows.(1) A new multi-services broadband access solution called HiNOC (High performance Network Over Coax) based on the out-of-band channels of CATV networks is proposed, aiming at promoting the growth of two-way transformation and broadband access technology of CATV networks. By means of out-of-band channels beyond 860MHz, HiNOC provides 40Mbps of MAC access rate within a single channel bandwidth of 16MHz in the last 100 meters'coax distribution network, and multiple services including streaming media are supported. Based on the analysis of channel characteristics and user requirements, the research objective and networking schemes of HiNOC are presented. The topological structure, duplex mode and multiple access mechanism of HiNOC MAC (HMAC) are addressed. With a star topology coordinated by a master node, a flexible two-way bandwidth allocation using TDD as well as a fully coordinated TDM/TDMA mechanism based on the reservation/grant mode, HMAC realizes a channel access plan with no collision and a dynamic resources allocation. It provides a new and viable solution for the network convergence.(2) HMAC protocol is designed and a thorough research is made on it. Firstly, a layered model of HiNOC is discussed. Secondly, CPS (Common Part Sublayer) and CS (Convergence Sublayer) of HMAC are designed in detail, and their mechanisms are studied in depth. CPS fulfills the core functions of HMAC, including node admission, channel reservation/allocation, link maintenance as well as QoS guarantee based on the priority, while CS fulfills the adaptation between the upper layer and HMAC core layer, including packing/unpacking, address learning as well as frame forwarding. Thirdly, the effect of parameters of HMAC protocol on the performance of HiNOC is analyzed. Finally, the experiments on the performance of HMAC are performed by means of HMAC testing subsystem. The experiments indicate that HMAC shows good transmission efficiency and throughput as well as steady transmission delay.(3) The single-rate multicast congestion control for streaming media applications is studied and two algorithms are presented. Firstly, a receiver-driven unicast congestion control algorithm called binomial congestion control algorithm at receiver (BAR) is proposed. In BAR, the receiver adjusts its congestion window based on the TCP-friendly binomial algorithm, and converts it to its own expected rate. The rate is then fed back to the sender who updates the sending rate accordingly. The simulation results show BAR's TCP-friendliness and excellent smoothness in rate fluctuations. Based on BAR, a single rate multicast congestion control for streaming media applications called the binomial congestion control at receivers for multicast (BARM) is proposed. To improve the scalability, the multicast receivers complete most tasks of congestion control. Each receiver maintains a congestion window independently, and adopts the binomial algorithm to adjust the window size from which an expected sending rate is calculated. To suppress feedback implosions caused by potentially large numbers of receivers, a representative selection policy and a distributed feedback suppression mechanism are used. Simulations indicate that BARM shows good TCP-friendliness, smoothness, scalability, and acceptable responsiveness.(4) A window-based layered multicast congestion control (WLMCC) algorithm is proposed. In WLMCC, the sender uses a set of multicast groups (called layers) to send data, and adjusts the sending rate of each layer adaptively according to the feedbacks of the receivers. Each receiver can individually select to join as many layers as permitted by the congestion condition between that receiver and the sender, thus acquiring a corresponding throughput. To detect congestion, each receiver independently maintains a congestion window, which is adjusted using GAIMD algorithm. According to the current congestion window an expected rate is calculated, based on which the receiver then cumulatively subscribes to appropriate layers. To prevent the problems of feedback implosion and drop-to-zero caused by LPM, a representative-based feedback suppression policy is employed on each layer. Simulations indicate that WLMCC provides different sending rate for each receiver experiencing different congestion conditions, thus improving the scalability of the WLMCC. Moreover, it shows good TCP-friendliness, smoothness and responsiveness.