软件定义网络中多媒体传输路由及缓存算法研究

发布时间：2018-01-01 21:52

本文关键词：软件定义网络中多媒体传输路由及缓存算法研究　出处：《中国科学技术大学》2016年博士论文　论文类型：学位论文

【摘要】：随着多媒体技术、计算机网络技术以及人机交互技术的飞速发展,多媒体内容开始逐渐深人到人们的日常交流、工作以及娱乐等各个方面,而由其所带动的多媒体应用也开始改变人们使用因特网的方式,将人们对高质量多媒体内容的需求带入了前所未见的高度。现如今,人们利用社交网络分享的内容已经不仅仅局限于文字和图片,音频和视频内容也变得非常常见。而近年来逐渐流行的高分辨率设备以及虚拟现实等新兴技术,在丰富了用户感官体验的同时,也显著推动了人们对高质量视频内容的需求。作为这一发展的必然结果,经由网络传输的多媒体内容,无论是在数量、大小还是类型上均有显著增长。飞速增长的多媒体流量给网络传输带来了包括服务质量保证、传输效率以及服务异构用户等在内的多项挑战,不同领域的研究人员也分别从编码、网络架构以及传输方案等角度提出了相应的改进措施。其中,对网络架构进行改进有着特殊的意义。一方面,现在的互联网架构是在上世纪末基于点对点通信模型而设计的,在面对如今大量应用所需求的一对多服务模型时,不能提供很好的支持。另一方面,优化底层网络架构,可以从根本上解决传统网络在承载视频业务时遇到的瓶颈,为新型编码方案(如分层编码等)以及新型传输方案(如自适应串流等)提供了实际应用的基础。例如近年来流行的软件定义网络(software-defined networking, SDN)和命名数据网络(named data networking, NDN),分别从灵活路由管控和一对多分发这两方面出发,针对传统网络在传输多媒体内容时的固有缺陷,提出了有效并且根本的解决方案,为提高多媒体内容的网络传输性能创造了极大潜力。本文主要以目前学术界热门的软件定义网络架构为基础,考虑多媒体内容传输的性能和效率优化。我们将考虑经典软件定义网络架构下的路由问题,以及带有缓存节点的软件定义网络架构下的路由和缓存问题。具体来说,我们首先考虑传统无缓存功能的软件定义网络架构,在传输多媒体内容时,利用多路径路由算法满足多媒体内容对带宽、延时、路径差分延时等指标的要求,并配合多媒体业务的分层特性来提高接收用户的体验。在此基础上,我们继续考虑支持网络中缓存的软件定义网络架构,希望通过在部分网络节点引入缓存功能来提高网络的传输效率。在支持网络中缓存的软件定义网络架构下,我们通过设计新的路由算法,在计算路径时考虑利用缓存节点来提高网络的传输效率。最后,在优化路由的基础上,我们展开对缓存节点的优化,利用在线机器学习技术设计热度感知缓存替换算法,显著提高缓存节点的缓存命中率,从而协助路由算法进一步节省带宽占用。论文的主要创新点以及贡献如下：1)本文提出了一种针对分层多媒体业务的多路径传输方法,通过在分配路径时考虑多媒体内容的层间优先级.为高优先级数据分配质量更好的链路,从而针对性地提高分层多媒体业务的服务质量,为分层编码技术提供了应用土壤。我们所提出的路由算法可以同时考虑用户请求多媒体内容时对带宽、延时、路径间差分延时以及层间优先级的要求,利用软件定义网络的灵活路由特性,为用户请求动态计算并分配相应的路径,从而提高网络传输性能。2)本文提出了一种支持网络中缓存的路由方法,为了充分利用网络中具有缓存功能的中间节点,我们设计了一套基于整数线性规划(integer linear programming, ILP)的路由算法,利用软件定义网络的全局路由特性,将流量合理分配、汇聚至缓存节点,以充分发挥缓存节点对带宽的节省作用,从而提高网络传输效率。3)本文提出了一种针对多媒体内容的网络缓存技术,我们设计了一套基于在线机器学习的新型缓存替换算法。算法通过学习多媒体内容在何种情况下热度如何变化,从而对内容的未来热度作出精确预测。相比于直接学习内容的热度,我们的方案有着更快的学习速度。由于采用无模型估计,我们的方案也打破了传统算法对内容热度分布函数的依赖,并且能快速追踪内容热度随时间的变化。我们从理论上证明了算法学习性能收敛至最优,也证明了缓存命中率次线性收敛至最优。在使用了真实系统数据的仿真评估中,我们所提出的算法不仅在缓存命中率上优于对比算法,并且也大幅降低了缓存替换次数,有效缓解了传统缓存替换算法对网络缓存系统产生过大I/O压力这一问题。我们所提出的算法有着对数时间复杂度,实际运行速度快于LFU等工业界常见算法。
[Abstract]:With the rapid development of computer multimedia technology, network technology and human-computer interaction technology, multimedia content gradually deep into people's daily communication, each work and entertainment, and by the multimedia application to begin to change how people use the Internet, the demand of high quality multimedia content into the unprecedented height. Nowadays, people use social networks to share content is not limited to text and pictures, audio and video content has become very common. High resolution devices gradually popular in recent years, virtual reality and other emerging technologies, in the rich user sensory experience at the same time, also significantly promoted the demand for high quality video content. As the inevitable result of the development of the multimedia content through the network transmission, whether in quantity, size or class Type had significant growth. The rapid growth of multimedia traffic brought include quality of service to network transmission, transmission efficiency and service of heterogeneous users including a number of challenges, researchers in different fields were also from the encoding, network architecture and transmission scheme and corresponding improvement measures were put forward. The structure of the network is a special significance improvement. On the one hand, the current Internet architecture is at the end of last century and based on the design of point-to-point communication model, in the face of today's application needs a lot of service model, can not provide good support. On the other hand, the optimization of the underlying network architecture, can fundamentally solve the the traditional network bottlenecks encountered in carrying video services, as a new encoding scheme (such as hierarchical encoding etc.) and a new transmission scheme (such as adaptive streaming etc.) provides a practical application For example, popular in recent years. Based on Software Defined Network (software-defined, networking, SDN) and data network (named data named networking, NDN, respectively) from the flexible routing control and the distribution of the two aspects, aiming at the inherent defects in the traditional network transmission of multimedia content, put forward effective solution and fundamental. Create a great potential for improving the performance of network transmission of multimedia content. This paper mainly to the current academic circles popular software defined network architecture, performance and efficiency of the optimization of multimedia content transmission. We will consider the routing problem of classical software defined network architecture, and software defined network architecture with the routing and cache node cache problem. Specifically, we first consider the software defined network architecture without traditional cache function, in the transmission of multimedia content, and Multi path routing algorithm to meet the multimedia content of the bandwidth, delay, delay path difference index requirements, and with hierarchical characteristics of multimedia services to improve the user experience. On this basis, we continue to consider the software defined network support network cache, in the hope that through part of the network nodes is introduced to improve the cache function the transmission efficiency of the network. In the software defined network support network cache, we design a new routing algorithm, in consideration when calculating the path to improve the efficiency of the transmission network using the cache node. Finally, on the basis of route optimization, we carried out optimization of cache nodes, using online machine learning technology design. Aware cache replacement algorithm, significantly improved cache node cache hit rate, so as to assist the routing algorithm further save bandwidth on the main. Innovations and contributions are as follows: 1) this paper proposes a multi path transmission method for layered multimedia services, by considering the multimedia content in the distribution path between the layers of priority. To link the high priority data distribution quality better, so as to improve the hierarchical quality of multimedia services, and provides the application of layered soil encoding technology. Our proposed routing algorithm can simultaneously consider the user request multimedia content on bandwidth, delay, delay and divided layer priority difference between path, using flexible routing characteristics of software defined network, users request dynamic calculation and distribution of the corresponding path, so as to improve the transmission performance of the network.2) is proposed in this paper. A routing method for caching support network, in order to make full use of the intermediate node has a buffer function in the network, we design a set based on the whole The number of linear programming (integer linear programming ILP) routing algorithm, using the global routing characteristics of software defined network, the flow converges to the rational allocation of cache nodes, in order to give full play to the role of the cache node to save bandwidth, improve network transmission efficiency of.3) this paper proposes a web caching technology for multimedia content. We designed a set of online machine learning based on the new cache replacement algorithm. The algorithm through how to change heat of the circumstances in which the multimedia content on the content of learning, to make accurate prediction of future heat. Compared to the direct learning content of the heat, our scheme has faster learning speed. By adopting the model estimation, we the scheme has also broken the traditional algorithm depends on the content of heat distribution function, and can quickly track the heat content changes over time. We have theoretically. It is proved that the algorithm converges to the optimal learning performance, also proved that the cache hit rate of linear converge to the optimal. In the use of real data in the simulation and evaluation system, our proposed algorithm not only in the cache hit rate is better than the comparison algorithm, and also significantly reduce the number of cache replacement, effectively alleviate the traditional cache replacement algorithm excessive pressure on the I/O network caching system. The proposed algorithm has a logarithmic time complexity, the actual operating speed faster than the LFU industry common algorithm.

【学位授予单位】：中国科学技术大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TN919.8

【相似文献】