Font Size: a A A

Research On Transmission Mechanisms For Streaming Media In Heterogeneous Networks

Posted on:2008-02-18Degree:DoctorType:Dissertation
Country:ChinaCandidate:M J LiuFull Text:PDF
GTID:1118360212999102Subject:Communication and Information System
Abstract/Summary:PDF Full Text Request
With the widespread penetration of broadband accesses, multimedia services are getting increasingly popular among users and have contributed to a significant amount of today's Internet traffic. However, there are still many problems in transmission mechanisms for streaming media, due to the Internet's best-effort service model and the inherent heterogeneous network environments. Focused on the transmission mechanisms for streaming media, this paper discussed and analyzed the problems of quality of service for multimedia transmission, congestion control in heterogeneous networks, application layer multicast and other related areas. The major contributions of this thesis are listed as follows:Streaming media requires Internet provide special quality of service (QoS) for its transmission, such as, delay, delay-jitter, bandwidth, and reliability. However, the Internet only provides the best-effort service, and no QoS guarantees provided. To satisfy different QoS requirements for streaming media, some mechanisms need to be employed. Among QoS guarantee mechanisms, QoS Routing is one of key issues. Chapter II researches some problems in QoS Routing, and presents a new Multi-constraints Active QoS Routing protocol, called MAR. MAR is a unicast QoS routing by applying Active Network technology into an improved distributed QoS Routing with selective probing. The task of MAR is to establish a path from the server to the client which has sufficient resources to support the required end-to-end QoS. Comparing with existed routing algorithms, MAR has three advantages: (a) the tentative paths are determined by the source node, so the communication overhead is little and has no problem of routing-loop and routing termination in MAR; (b) Each node only maintains local state information and the global topology of the network, thus MAR can avoid the global link state information exchange and reduce the communication overhead; (c) MAR can decrease the computing complexity by decentralizing the constraints judgments in intermediate nodes. Some simulations and analysis have proved MAR is more concise, more controllable, and more flexible than other QoS routing mechanisms.IP Multicast is the most efficient way to distribute streaming media form the server to a large number of receivers. For the heterogeneity of Internet, one of the significant remaining hurdles to widespread adoption of IP multicast is the development of suitable congestion control algorithms. State of art, Layered multicast is a promising solution to accomplish this task efficiently. Chapter III proposes a new Layered Multicast with Multilevel Congestion Marking (LM-MCM). A fundamental contribution of this scheme is introducing a Multilevel Congestion Marking strategy (MCM) which enables a router to mark the congestion-flag field of each arriving LM-MCM packet according to the average queue size. Furthermore, based on the average congestion-flag instead of loss rate, each receiver can improve/reduce the quality of service it receives to respond to the network congestion. On the other hand, MAR can emulate the behavior of TCP's congestion control mechanism by setting each layer's transmission rate and join-timer appropriately. We evaluated LM-MCM in a number of scenarios on NS2 simulator. The results show LM-MCM converges quickly to the optimal layer of subscription, induces less loss to track the available bandwidth, and has inter-session fairness and TCP-friendliness.LM-MCM can resolve Layered Multicast's problems perfectly; however its benefits are achieved by introducing a Multilevel Congestion Marking strategy (MCM) in routers, which limited the scalability of the system. Another alternative is to detect the network congestion by analyzing the arriving packet's delay-jitter trend in the receiver. In Chapter IV, we have discussed the relationship between the queuing delay and the packet's delay-jitter trend, and have a conclusion that each receiver can implicitly infer the network congestion at early by analyzing the base layer packet's delay-jitter trend in every synchronization period. Based on this conclusion, we proposed a new Layered Multicast based on Delay-Jitter Trend (LM-DJT). LM-DJT introduces the Delay-Jitter Trend model into an amended TCP throughput equation to estimate the TCP-friendly transmission rate, and each receiver adjusts its subscription layers according to the estimated transmission rate and an adaptive strategy of layer's subscription. Our simulation results demonstrate that, under the same network topology and condition, the LM-DJT architecture can achieve higher network throughput and smoother video qualities on the receiver side than the existing approaches.Recent years, Application Layer Multicast (ALM) has been presented as an alternative to IP Multicast. In ALM, end-systems (peers) participating in a streaming session self-organize into an overlay structure and media distribution is then achieved through data relaying among these peers. Because no support from network infrastructure is required, ALM is a promising approach for streaming media applications. In Chapter V, we study the application layer multicast with multi-senders; a fundamental problem of which is how to organize participating peers into a multi-sender based overlay. This chapter presents a simple and scalable approach for constructing DHT-Circle based Overlay network, called DHTCO. DHTCO organizes all participating peers into a hybrid two-layer structure, the lower-layer is a structured circle to maintain peer information by using a simplified distributed hash table, the upper-layer is an unstructured overlay over which each peer can find other active peers in the circle and determine its own sender peers independently. Furthermore, an algorithm is proposed to help each peer select its own sender peers adaptively. Our preliminary results demonstrate that DHTCO achieve quite good streaming quality under local network conditions. Moreover, its control overhead and transmission latency are both kept at acceptable levels.
Keywords/Search Tags:Streaming media, QoS Routing, Multi-rate Congestion Control, Peer-to-Peer Network, Application Layer Multicast with multi-senders
PDF Full Text Request
Related items