The Research On Network Coding In Multicast Information Flow

Since the ATM switching was placed by packet switching in TCP/IP protocols, there are rarely major revisions on transmission protocols for several decades. However, the situation has been changed by introducing a new technology called as network coding, which are presented by Ahlswede et al.. It brought a new age for information theory. With network coding, information can be distributed after integration in multicast scenarios, which noticeably increases transmission rates. With the leverage of network coding in multicasts, the subject of transmission is not how to distribute the information, but the capacity of network information flow. In the traditional schemes, information is delivered and processed in packets, which brings burdens due to redundancies, computing and modeling complexity. The burdens become huge when the network scale grows with large group of clients and the corresponding growing requirements on services. One of the approaches to lower the burdens is to transmit and process information by flow. As the main motivation of my research work, the questions rise about how to represent information flows and how to apply them.Network information flow in this thesis is not the information flow defined for ATM switching. Based on network information theory, it can be formed as multiple specifications in different functions, like vector for coding, such that the entropy of the flow is hardly defined by Shannon's theory. However, it is much simpler to measure, process and correct in big scaled network with much lower cost than packet switching. To apply information flow in real network, we should further consider the leverage with other theory, such as graph theory and queue theory.With the aforementioned conceptions, the thesis first presents the definition of network information flow and the core idea of network coding. I also discuss a new metric, which is called the degree of freedom, to measure the dimension of information flow field. Based on flows, I explored new approaches applying network coding to be compatitable with different multicast scenarios, as well as the corresponding revisions on transmission protocols. In theory, network coding can assists multicast transmissions to achieve their maximum flow, which are the upper-bounds of transmission rates. However, in practice, the upper-bounds are hard to achieve without specially designed transmission schemes, which are referred by theoretical results. The research interests and the corresponding contributions of this thesis are list as follows.First, I establish a linear space model for information flows and explain the maximum flow in multicast networks based on the model. Furthermore, I analyze the capacity of multicast networks, and show the advantages to apply network coding in multicast networks by comparing with the optimal results by fractional routing in traditional protocols. After the illustrations of various coding algorithms for network coding, I present a novel coding algorithm to employ flow as the coding unit and shuffle them for coding, which lowers the complexity for both encoding and decoding.Second, I studied the Head of Line blocking problem in network multicast switches and deduce the problem as the graph coloring problem. The function of network coding in multicast switches is presented as well as the proof on its promising enhancement. I further propose a new coding algorithm based on Vandermond matrices, which can lower the requirements of coding processes on switch buffer as well as the complexity.Third, based on the theoretical results in my second contribution, I further analyze the gap between theory and practice when applying network coding in frame-based multicast switches. To assist network coding in switches, I propose a so-called Coding-Driven scheduling algorithm instead of traditional scheduling policies, and enhance performances of switches with network coding.Fourth, under the mode of cognitive relay with multiple assistants,I aim to handle the lost information when propagating through wireless medium. With the character of wireless broadcasting in the spectrum, the information flows go through different sensing channels synchronically. A further estimation on the entropy of end-to-end information flows is made in the model according to CSI (Channel State Information). Accordingly, with network coding, I propose a so-called CodeAssist relay protocol, which can decrease the redundancy of relay and enhance the efficiency to use the spectrum.According to analysis on network properties, the promise of network coding on network transmission is given in this thesis. By establishing universal models in different applications, I analyze the properties of information flows and revise the correlated transmission schemes to enhance the efficiency. I also support the information flow theory by empirical results. It is demonstrated in both theoretical and empirical phases that, to model and design transmission protocols for network multicast missions with integrated information flows will benefit the leverage of network coding in practice. It will definitely decrease the redundancies in transmissions and enhance the transmission efficiency.