提高无线机会性网络编码性能增益机制研究

发布时间：2018-03-27 15:31

  本文选题：机会性网络编码　切入点：无线网络　出处：《电子科技大学》2014年博士论文

【摘要】：网络编码(Network Coding)被视为一种有前景的技术,在无线网络中可用来减少传输次数,提高带宽利用率,改善网络吞吐率和能量效率。如何设计更好的网络编码架构和机制来充分发挥网络编码的性能增益是目前学术界的研究热点。网络编码的基本思想是允许网络中间节点将多个数据包融合为一个编码数据包加以发送,从而减少传输次数,改善网络吞吐率。受益于无线通信内在的广播特性和侦听能力,网络编码更加适合应用于无线网络。为使得网络编码在无线网络中应用时尽可能减少对网络协议栈的改变,机会性网络编码是一个可行的选择。根据机会性网络编码不同的应用场景,需要考虑多种因素对其性能增益的影响,比如节点的缓存资源、处理能力等。针对不同的因素,建立相应的数学模型,优化其在无线网络中的性能增益。因此,针对机会性网络编码在无线网络中的应用,本文指出了在不同网络场景和情形下需要进一步研究的问题,并提出了相应的优化架构和机制。全文的主要内容包括:首先考虑缓存资源对机会性网络编码性能增益的影响。现有机会性网络编码架构都假定节点拥有无限的缓存资源和处理能力。在此前提下,网络各节点需要解码包池缓存所有侦听到的和已发送的数据包足够长时间,用以未来可能的解码需要。同时,节点间通过“接收报告”相互周期性地交换各自包池中所拥有数据包信息,以确保编码包的可解性。然而,当节点缓存空间受限时,机会性网络编码的性能增益将会受到影响。即使节点资源无限大,现有的机会性网络编码解码包池缓存机制会带来更大的网络开销、从而降低网络吞吐率。为了解决这一问题,本文提出了一种机会性网络编码框架来优化缓存资源的解码效用。通过分析,我们将缓存资源与解码性能之间的关系归结为一个最优化数学模型。通过这个优化模型,我们推导出一系列的缓存规则,并提出了一种分布式的机会性网络编码缓存策略。仿真结果证实了该模型的有效性,并证明该策略可以有效改善节点缓存资源利用率,提高机会性网络编码性能增益。对于机会性网络编码在受干扰无线网络中的应用,现有机会性网络编码架构通过伪广播一次性将一个编码包投递至多个接收节点,并依赖于捎带在数据包包头中的异步ACKs来确认接收。在给定时间内未被确认接收的数据包将被再次插入发送队列、编码发送。这种机制在丢包无线网络中容易由于数据包的丢失和延迟到达而招致大量冗余重传、浪费网络带宽。此外,为了确保编码包的可靠投递,现有网络编码架构默认编码节点采用最低传输速率来加以投递。这样会导致较长的传输时间,并增加了数据包发生碰撞的可能。针对以上问题,我们提出了一个结合传输速率控制和网络编码码字选择的框架ORC来对网络层和MAC层进行联合优化。该框架将编码包传输的速率控制问题归结为一个马尔科夫决策过程,从而获得最优的速率选择策略。然后,基于以上速率选择后的结果,从所有可能的编码包组合中选择最佳的编码包组合并予以发送。对于编码包组合问题,我们将其归结为一个最大加权团问题,并提出了一个码的选择算法。网络仿真证实,框架ORC可以有效改善网络吞吐率,降低端到端的投递延迟。现有机会性网络编码架构中存在的另一个问题是编码机会不足。由于无线介质访问的随机性和各条编码数据流速率的不匹配,可能导致编码机会不足,进而降低了编码增益。主动性延迟数据包发送可以增加编码机会,但是会增加投递延迟,对实时应用尤为不利。为此,我们提出了一种结合调度和网络编码的框架SNC,在有延迟限制条件下最大化编码机会。该框架由延迟数据包传输策略DTP和基于网络编码组的调度策略GSP两部分组成。前者DTP基于延迟要求和可以取得的最大编码度,动态地调整可以取得的编码度,从而最大化系统中的编码机会。在此基础上,对于每一个网络编码组,调度策略GSP依据权重来调度网络编码组对应的编码包发送,从而优化机会性网络编码吞吐率增益。网络和数值仿真证实,SNC可以最大化编码机会,改善机会性网络编码在实时应用中的吞吐率性能增益。尽管主动性推迟数据包的发送可以增加编码机会,现有的工作往往局限于特定的网络拓扑并具有较高的计算复杂度,不易于分布式实现。为此,我们提出了具有编码意识的队列管理机制DQM,充分开发了网络编码在存储压缩上的优势,进一步增加了编码机会。仿真结果证实了DQM的有效性。本文取得的研究成果,为推进网络编码技术在无线网络中的实用化提供了有用的理论基础和技术手段。
[Abstract]:Network encoding (Network Coding) is regarded as a promising technique in wireless networks can be used to reduce transmission times, improve the utilization rate of bandwidth, improve network throughput and energy efficiency. The network architecture and how to design a better encoding mechanism to give full play to the performance gain of network encoding is a hot topic in academic circles. The basic idea is to allow the network encoding network node multiple data packets into one encoding packets to be sent, so as to reduce transmission times, improve the network throughput. Benefit from wireless communication within the broadcast nature and listening ability, network encoding are more suitable for application in wireless network. In order to make the network as much as possible to reduce the encoding the network protocol stack changing applications in wireless networks, opportunistic network encoding is a feasible choice. According to different application scenarios of opportunistic network encoding, need Considering the influence of various factors on the performance gain, such as cache resource node, processing ability. According to different factors, establish the corresponding mathematical model and its optimization in wireless network performance gain. Therefore, the application of opportunistic network encoding in wireless network, this paper points out the problems that need further study in different network scene and situation, and put forward the corresponding optimization framework and mechanism. The main contents of this paper include: firstly, considering the influence on the performance of opportunistic network cache resource encoding gain. Encoding architecture of existing opportunistic network are assumed to have infinite node buffer resources and processing ability. Under this premise, the network nodes need to decode the packet buffer pool all detected and sent enough packets to decode the long time, possible future needs. At the same time, nodes through "receiving report" Periodically exchange with packet information of each packet in the pool, to ensure the solvability of the encoding package. However, when the cache space is limited, the performance gain will be affected by the opportunistic network encoding. Even if the node resource is infinite, the existing network encoding decoding opportunistic packet pool cache will bring greater network overhead, thereby reducing the throughput of the network. In order to solve this problem, this paper proposes a decoding utility framework to optimize the opportunistic network encoding cache resources. Through the analysis, we will shut down between the cache resources and decoding performance as an optimization mathematical model. Through this model, we derive a cache rule series, and proposed opportunistic network caching scheme for distributed encoding. The simulation results confirm the validity of the model, and it is proved that this method can effectively improve the node The utilization of cache resources, improve the performance of opportunistic network encoding gain. For opportunistic network encoding interference in the application of wireless networks, the existing network architecture through opportunistic encoding pseudo broadcast once a encoding packet delivery to multiple receiving nodes, and relies on the piggyback asynchronous ACKs in the header of each packet to confirm receiving packets in. In a given period of time is not confirmed will be received again into the send queue, encoding sent. This mechanism in packet loss in wireless network easily due to packet loss and delay to incur a lot of redundant retransmission, the waste of network bandwidth. In addition, in order to ensure the reliable delivery of packet encoding, the existing network architecture with the lowest node encoding default encoding the transmission rate to be delivered. This will lead to a longer transmission time, and increase the packet collision. To solve the above problems, I have Put forward a combined transmission rate control and network encoding codeword selection framework ORC to the network layer and the MAC layer joint optimization. The framework will rate control problem encoding packet transmission is considered as a Markov decision process, so as to obtain the optimal rate selection strategy. Then, after the above rate selection based on the results. From all the possible combinations to select the best encoding packet encoding and sending packet combination. For encoding a package, we will be attributed to a maximum weighted clique problem, and propose a code selection algorithm. The network simulation confirmed that the ORC framework can effectively improve the network throughput and reduce the end to end the delivery delay. Another problem existing in the opportunistic network encoding architecture is lack of opportunities. Due to the mismatch encoding wireless medium access randomness and the encoding rate of flow of data, can Lead to the lack of opportunities and reduce the encoding, the encoding gain. Active packet delay can increase the encoding opportunity, but will increase the delivery delay, is particularly detrimental to the real-time application. Therefore, we propose a SNC framework combined with the scheduling and network encoding, the delay constraint conditions to maximize opportunities. The framework consists of encoding delay of data packet transmission strategies for DTP and GSP scheduling strategy of network encoding group is composed of two parts. Based on the former DTP based on delay requirement and can achieve the maximum degree of encoding, dynamic adjustment can be made so as to maximize the system of encoding, the encoding opportunity. On this basis, for each network encoding group, scheduling strategy GSP based on the weights of network scheduling encoding groups corresponding to the encoding packets, so as to optimize the opportunistic network encoding throughput gain. That network and numerical simulation, SNC can maximize the encoding machine Will improve the throughput performance gain of opportunistic network encoding in real-time applications. Although the initiative to postpone sending data packets can increase the encoding opportunity, existing work is often limited to specific network topology and the computational complexity is higher, not easy to realize distributed. Therefore, we propose a DQM queue management mechanism with encoding consciousness the full development of the network in the compressed storage encoding advantage, to further increase the encoding opportunity. The simulation results verify the effectiveness of DQM. The result of this research, provides the theory basis and the technical means for promoting the practical useful network encoding technology in wireless network.

【学位授予单位】：电子科技大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TN92
，

本文编号：1672121