智慧协同网络资源动态适配机制研究

发布时间：2018-04-20 01:32

本文选题：资源适配 + 族群　；参考：《北京交通大学》2016年博士论文

【摘要】：服务需求与网络资源间的匹配是互联网发展各个阶段都在努力追求的目标。传统互联网架构在资源适配问题上存在局限性。在传统网络基础上的改进方案效果依然不甚理想。近些年,国内外学者开始研究面向资源、内容、服务的各种未来互联网架构。智慧协同网络架构作为一种面向资源的未来互联网解决方案,创造性地提出了资源适配层的概念。资源适配层通过族群的智慧博弈决策来实现服务需求和网络资源之间的匹配,实现智能服务分发和网络资源高效利用。本论文重点围绕资源适配层如何通过族群博弈决策机制来实现流量调配和内容分发,以及如何通过多路径协同适配机制来实现智慧传输控制展开深入研究。本论文具体完成以下工作：(1)在服务层与网络层跨层协作机制方面,研究了应用层自私路由和网络层流量工程之间的相互影响。通过构建两种非合作动态博弈模型来分别描述应用层的自私拓扑感知行为和网络层的自私流量监测行为,通过理论分析得出当应用层或网络流量工程单方面实施自我优化策略时,得到比双方非合作博弈时更少的收益。由此获得启发构建一种两层之间协作的博弈模型,提出一种基于信息交互的跨层协作开销节省算法,实现服务层用户就近获取资源,同时网络传送流量的开销得到优化。(2)在资源适配层族群博弈决策机制方面,族群博弈决策的智慧性在于能根据不断变化的服务需求和网络资源状态对适配策略做出动态的调整；它的合理性在于博弈的纳什均衡点处的策略使各方利益最大化,改变策略不能获得更大利益。本文提出一种满足服务需求的族群内协作队列模型,在此基础上提出族群博弈模型,博弈均衡点处使时延敏感服务和丢包敏感服务两种服务质量得到最优化,同时使网络节点维持负载均衡。(3)在族群间博弈决策方面,提出了四种博弈模型来分析族群间基于内容分发的相互关系。智慧协同网络的资源适配在规模上应该具有可扩展性。这需要区域族群之间协作,实现跨族群资源适配。针对内容资源的跨族群分发问题,各族群通过参与博弈来决定对自己最优的内容缓存和内容转发共享策略。通过对四种模型下族群内容分发开销和分布式优化失效程度的比较分析,证明族群间协作模式的优势。(4)在传输控制机制方面,提出了满足服务质量需求的公平有效的多路径协同传输控制机制。面向内容资源的网络应该支持高质量服务(如多路径传输的媒体流等)。本文通过基于路径质量智慧感知的多路径传输控制,来实现多路径服务流的高效传输和对单路径服务流的友好性,从而使各种服务流在共享瓶颈链路资源时,达到公平的带宽资源分配,同时能够改善服务传输质量和满足用户体验需求。本论文围绕智慧协同网络架构下资源适配层的协作博弈决策和智慧控制机制来展开思路,研究了互联网服务层和网络层的跨层协作机制、族群内博弈决策机制、族群间博弈决策机制和多路径协同适配的智慧传输控制机制四部分内容,对智慧协同网络资源适配层理论体系的构建做出了一定贡献,具有一定的创新性。在研究方法上,本文运用了文献调研、模型构建、算法设计、数学推导、仿真评估等一系列研究方法,来实现研究目标。
[Abstract]:The matching between service requirements and network resources is the goal of the Internet development at all stages. The traditional Internet architecture has limitations on resource adaptation. The effect of the improved scheme on the traditional network is still not ideal. In recent years, scholars at home and abroad have begun to study various kinds of resources, content, and services. The intelligent collaborative network architecture, as a resource oriented future Internet solution, creatively proposes the concept of resource adaptation layer. The resource adaptation layer realizes the matching between service demand and network resources through the intelligence game decision of the ethnic group, and realizes the intelligent service distribution and the efficient use of network resources. This paper focuses on how resource adaptation layer can realize flow allocation and content distribution through ethnic group game decision-making mechanism, and how to realize intelligent transmission control through multi path coordination adaptation mechanism. The following work is completed in this paper: (1) in the service layer and the network layer cooperation mechanism, the research should be done. With the interaction between layer selfish routing and network layer traffic engineering, two kinds of non cooperative dynamic game models are constructed to describe the selfish topology perception behavior of the application layer and the selfishness monitoring behavior of the network layer. The theoretical analysis shows that when the application layer or the network flow project is implemented in a single aspect, the self optimization strategy is obtained. It is less profitable than the non cooperative game between the two parties. Thus, a game model between two layers is inspired to construct a game model of cooperation between layers, and a cross layer cooperation overhead saving algorithm based on information interaction is proposed to achieve the close access to resources by the service layer users, while the overhead of network transport traffic is optimized. (2) game decision making in the resource adaptation layer group. In the aspect of mechanism, the wisdom of ethnic group game decision is that it can adjust the adaptation strategy dynamically according to the changing service demand and the state of network resources. Its rationality lies in the strategy of Nash equilibrium point of the game to maximize the interests of all parties and change the strategy can not gain greater interests. This paper proposes a kind of service needs to meet the needs of service. On the basis of this, an ethnic group game model is proposed. On this basis, an ethnic group game model is proposed. Two kinds of service quality are optimized for time delay sensitive service and packet loss sensitive service at the equilibrium point. (3) four game models are proposed to analyze inter ethnic based game decision-making in inter group game decision-making. The resource adaptation of the cooperative network should be extensible on the scale. This requires cooperation between ethnic groups to realize the adaptation of inter ethnic resources. In view of the problem of inter ethnic distribution of content resources, each ethnic group determines the optimal content caching and content forwarding sharing strategy through participation game. The comparative analysis of the content distribution overhead and the degree of distributed optimization failure under the four models proves the advantages of the inter ethnic collaboration model. (4) a fair and effective multi-path cooperative transmission control mechanism to meet the service quality requirements is proposed in the transmission control mechanism. The content oriented network should support the high quality service (such as the content resource oriented network). In this paper, multipath transmission control based on path quality intelligent perception is used to achieve efficient transmission of multipath service flow and the friendliness of single path service flow, so that all service streams can share the fair bandwidth allocation when sharing bottleneck links and improve service transmission. This thesis focuses on the cooperative game decision and the intelligent control mechanism of the resource adaptation layer under the intelligent cooperative network architecture, and studies the cross layer cooperation mechanism of the Internet service layer and the network layer, the game decision-making mechanism in the ethnic group, the game decision-making mechanism between the ethnic groups and the wisdom of the multi-path coordination adaptation. The four part of the transmission control mechanism makes a certain contribution to the construction of the intelligent network resource adaptation layer theory system. In the research method, this paper uses a series of research methods, such as literature research, model construction, algorithm design, mathematical deduction, simulation evaluation and so on, to achieve the research goal.

【学位授予单位】：北京交通大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP393.02

【参考文献】