| Many applications based on Internet need to use the large-scale data delivery mechanism, such as IP multicast and P2P cooperative content distribution, to deliver information to a large number of users. How to obtain the better delivery performance with the limited network resources has become a"hot spot"in the research community of network information flow. To achieve the goal, network coding, as one of the most important breakthroughs on the theory of information processing and transmission, was proposed in recent years. The main principle behind network coding is that the relay (intermediate) nodes in the communication network make encoding and decoding operations to the data bits. Network coding generalizes the traditional data delivery mechanism where the data bits can only be stored and forwarded. With network coding,the communication network can achieve better transmission performance, such as higher network throughput and lower bandwidth consumption etc. Due to its significant potentials, network coding has received much concern in both the research and industrial community. Lots of famous universities and research agencies have put much attention on the research of network coding and finished some important works.It can be proved from the theoretical aspects that network coding can enhance the performance of a given data network to a large extent. However, compared with the traditional data delivery mechanism based on storing and forwarding, the nodes in network coding system need to perform encoding and decoding operations such that the system based on network coding takes additional computational overheads (such as I/O and CPU consumption), which will bring high cost (computational and non-computational) to deploy network coding in real data networks. Hence, it is better to find out an optimal network coding scheme in which the additional computational overheads caused by network coding is as lower as possible. To address this issue, this thesis investigates the optimization technologies of network coding from the theoretical and practical interests respectively. The content of this thesis includes three aspects: 1) Information flow vector-based optimal network coding model; 2) Key links-based low-cost network coding algorithm and its distributed implementation; 3) Clustered network coding-based cooperative content distribution scheme and algorithms.The works in this thesis have been supported by the National Science Foundation of China"Wireless Mesh Network-based Stream Media Distribution"(No. 60773193), High-Tech Foundation of Huawei Corporation (China)"Peer to Peer Oriented Network Coding"(No.YJCB2006049RE), the Foundation of Hubei Provincial Key Laboratory of Smart Internet Technology (HSIT-200605) and the Research Foundation of Department of Electronics and Information Engineering, Huazhong University of Science and Technology (2008).The contributions of this thesis contain:1. Information flow vector-base optimal network coding model. By modeling the information transmitted in the data delivery network as network information flow, we present a simple but effective approach to describe the information flow on each link in the data delivery system, referred to information flow vector. We analyze the properties of information flow vector, and conclude some requirements the information flow vector should meet with, and then construct an optimal network coding model. With the model, the problem of reducing the additional computational overheads of network coding operations can be converted into a programming formulation. Therefore, we can find out the optimal network coding-based data delivery scheme with minimal computational overheads by solving the model. Moreover, we consider some special cases and proposed a utility tradeoff-based network coding model in which the utility and computational overheads are considered together.2. Key link-based low-cost network coding algorithm and its distributed implementation. From the viewpoint of practical interests, we analyze the inherent characteristics of network coding and show that the essence of network coding is that there are some key links shared by deferent paths between the source and multi-receivers when we establish the transmission paths for the data delivery network. To achieve the maximal throughput of the network, network coding is needed and the key links should carry encoded information. Since the additional computational overheads in network coding system are caused by the network coding operations, an effective way of reducing additional computational overheads and realizing a low-cost network coding transmission scheme is to reduce the number of key links while constructing transmission paths. Following this idea, we present a key links-based algorithm for low-cost network coding and propose its distribution implementation.3. Cluster network coding-based content distribution scheme. Due to the technical and non-technical reasons, the IP multicast based data delivery system which the network coding original proposed for has not been widely implemented in the network till now. As a scable solution to deliver content to a large number of end-users, the overlay network based P2P cooperative data delivery scheme and systems have been widely used recently. Borrowing the idea of network coding, a novel data delivery scheme which is based on network coding was proposed. However, some reaserchers think the performance of such novel scheme be less optimistics due to the additional computational overheads caused by network coding operations. To optimize the data delivery scheme and enhance the performance further, we proposed some new mechanisms, such as clustered network coding, random choice-based strategy and independence checking algorithm etc. Based on the proposed mechanisms, we construct an improved network coding-based data delivery scheme. Compare to the common used data delivery scheme, the improved scheme can reduce the additional computational consumptions of network coding operations to a large extent, and improve the performance of data delivery system significantly.
|
PDF Full Text Request