Application of Network Coding in Emerging Communication Networks

线性网络编码是最近为改进网络信息传输而开发的技术。从一般的角度来看，来自各种来源并去往不同接收器的数据分组不是逐个分组地传输的；相反，生成和发送分组组的多个线性组合。通常，这样的线性组合(编码的分组)涉及与随机系数的乘法，并且由充当源和目的地之间的转发中继的中间节点来实现。正确获取所需数量的线性独立分组组合的接收器能够解码并保留发往它的分组。目前的研究表明，网络编码可以通过增加数据吞吐量和对网络攻击的弹性来改善网络性能。然而，网络编码确实会导致网络元素上的计算压力增加(这个问题可以使用现代功能强大的处理器来解决)，以及增加网络的复杂性和可扩展性问题。直到最近，网络和使用它们的应用程序之间一直存在着明确的分离。应用程序生成通过简单套接字API发送的包；网络传递它们。然而，在许多情况下，应用程序可以从与网络更直接的交互中受益；例如，可以更多地观察当前网络状态，并获得对网络行为的更多控制。软件定义网络(SDN)定义了一种网络体系结构，该网络体系结构可以提供对网络控制平面的访问/可编程性，从而实现网络编码。*建议的研究将集中于网络编码在以下领域的应用：a)在无线网络(具有不可靠链路)以及软件定义网络(SDN)中广播/组播信息；b)在大型无线传感器网络(WSN)中集成网络编码和压缩感知技术。研究的重点将主要集中在网络性能、网络稳定性的表征、优化某些性能度量的最优随机控制(与网络吞吐量和延迟有关)以及使用网络编码的数据系统与现有网络中使用的传统技术之间的性能比较。我不会把重点放在优化网络编码的开发上(将假设为简单的随机线性编码)。*尽管最近网络编码在研究界受到了热烈的欢迎，但它还没有被工业产品开发阶段采用。因此，它是一个很好的研究领域，具有很好的实用价值和不久的部署前景。我坚信这不仅是HQP培训的绝佳领域，而且对加拿大经济特别重要。**

Linear network coding is a recent technique developed for improved network information transmission. From a general perspective, data packets originating from various sources and destined to different receivers are not transmitted on a packet by packet basis; instead multiple linear combinations of groups of packets are generated and sent. Typically, such linear combinations (encoded packets) involve multiplication with random coefficients and are implemented by intermediate nodes that act as forwarding relays between the sources and destinations. A receiver that correctly acquires a necessary number of linearly independent packet combinations is able to decode and keep the packets that were destined to it. Current research has shown that network coding may lead to improved network performance by means of increased data throughput and resilience to network attacks. However, network coding does lead to increased computational stress on network elements (an issue that can be resolved using modern powerful processors) as well as increased network complexity and scalability issues. ***Until recently there has been a clean separation between networks and the applications that use them. Applications produce packets that are sent over a simple socket API; the network delivers them. However, there are many occasions when applications can benefit from more direct interaction with the network; e.g., to observe more of the current network state and to obtain more control over the network behavior. Software Defined Networking (SDN) defines a network architecture that can provide access/programmability to the network control plane and thus enable the implementation of network coding.  ***The proposed research will concentrate on application of network coding in the following areas: a) broadcasting /multicasting information in wireless networks (with unreliable links) as well as in Software Defined Networks (SDN) and b) integration of network coding and compressive sensing in large wireless sensor networks (WSN). The focus of the research will be primarily on the networking performance; characterization of network stability, optimal stochastic control for optimizing certain performance measures (related to network throughput and delay) and comparison of performances between data systems using network coding versus traditional techniques used in existing networks. I will not focus on the development of optimized network codes (will assume simple random linear coding).***Despite the warm reception network coding has received in the research community recently, it has not been adopted at the industrial product development stage. Therefore, it is an excellent research area with promising practical implementations and expected deployment in the near future. I strongly believe that it is not only an excellent area for HQP training but also is of particular importance to the Canadian economy.**