智慧协同网络资源适配机制与方法研究

发布时间：2018-01-12 15:33

本文关键词：智慧协同网络资源适配机制与方法研究　出处：《北京交通大学》2016年博士论文　论文类型：学位论文

【摘要】：智慧协同网络(Smart Identifier NETwork/SINET)作为一种面向未来需求的新型网络架构,在有效解决可扩展性、移动性、安全性等问题的基础上,将“智慧性”内嵌于网络架构中,使网络能够原生地支持智慧型资源适配,从而提高资源利用率,降低能耗,从根本上实现高效节能的“绿色网络”目标。本文作为“智慧协同网络理论体系”的组成部分,围绕网络的“原生智慧性”,针对网络状态感知、资源动态调度、选择性休眠等机制与方法开展研究,主要工作和创新点如下：1.设计了一种基于状态感知的资源适配机制,作为智慧协同网络“原生智慧性”的实现基础。首先,在关键功能实体资源管理器和转发组件中嵌入感知功能,使转发组件能够实时感知并记录网络状态。其次,设计了一种智慧协同控制协议,供资源管理器控制转发组件执行基本的数据处理,以及获取网络状态信息。然后,提出了一种基于网络状态的路径计算方法。最后,在原型系统中验证关键功能实体和智慧协同控制协议的可行性,在仿真平台上对所提机制进行效果验证。结果表明,在智慧协同网络架构下运行该机制是可行的,并且该机制能够有效提高网络资源利用率,提升用户体验。2.设计了两种感知信息交互机制,即询问和汇报机制,作为资源适配机制的进一步细化完善。在询问机制中,资源管理器主动询问转发组件,得到静态的全局网络状态信息,该机制实时性强,但会带来较大的额外控制开销。在汇报机制中,转发组件主动推送动态的局部网络状态信息,该机制几乎不产生额外控制开销,但实时性受网络流量的影响。然后,在原型系统中分别实现两种机制,设计不同流量场景,验证比较两种机制的实时性。最后,利用真实流量数据,在开销及实时性两方面权衡折中,对两种机制的实际应用效果进行分析评估。结果表明,汇报机制的实时性不及询问机制,但汇报机制具备不产生额外控制开销的优势,虽然存在一定时延及误差,但其程度在可接受范围内,因此,利用汇报机制来传递感知信息是可行的。3.提出了一种基于流量感知的智慧路径分配方法,用以支持智慧协同网络的“资源动态调度”功能。首先,将实时测算出的流量矩阵作为约束,对负载均衡路由优化问题进行建模,再利用拉格朗日对偶方法,将原问题转化为对偶问题。然后,为求对偶问题最优解,设计了一种基于流量矩阵的负载均衡路径分配算法。最后,通过仿真实验和原型系统测试,验证所提算法的负载均衡效果及额外控制开销情况。结果表明,该方法能够有效避免拥塞,实现网络负载均衡,且其额外控制开销也在可接受范围内。4.提出了一种网络能耗和协作缓存优化方法,用以支持智慧协同网络的“选择性休眠”功能。智慧协同网络中的转发组件不仅执行简单的转发,还具备缓存及休眠功能,因此,为了在兼顾缓存性能的同时,最大化地节省网络能耗,建立了多目标混合整数线性规划问题模型。然后,将缓存性能优化目标转化为约束条件,使多目标问题转化为单目标问题,并设计休眠组件选择算法进行求解。最后,模拟真实网络流量,分别在三个真实网络拓扑中,对所提算法和现有算法的节能效果进行比较评估。结果表明：本文所提算法的节能效果要优于其他算法；并且网络的冗余链路越多,算法的节能效果越好；在网络闲时执行该方法,剩余网络仍然能保持较好的负载均衡水平。
[Abstract]:The wisdom of the collaborative network (Smart Identifier NETwork/SINET) is a new type of network architecture for future needs, mobility effectively solve the scalability, security and other issues, based on the "wisdom" is embedded in the network architecture, so that the network can support native intelligent resource adaptation, so as to improve the utilization rate of resources fundamentally, reduce energy consumption, to achieve energy-efficient "green network". This part as the "wisdom of the collaborative network theory system", "native wisdom around the network", the network state aware, dynamic resource scheduling, to carry out research on selective dormancy mechanism and method, the main work and innovation are as follows: 1. design an adaptive mechanism based on state aware resource, as the basis for network collaborative wisdom "native wisdom". First of all, a key function in the resource management entity For embedded sensing function and forwarding module, the forwarding assembly can be real-time sensing and recording network. Secondly, design a collaborative intelligence control protocol for resource manager control forwarding basic data processing component execution, and access to network information. Then, this paper presents a calculation method based on the state of the network path. Finally, verify the feasibility of key functional entities and collaborative intelligence control protocol in the prototype system, verify the effect of the proposed mechanism in the simulation platform. The results show that the mechanism in collaborative intelligence network architecture is feasible, and the mechanism can effectively improve the utilization rate of the two kinds of cyber source, sensing information exchange mechanism designed to enhance the user experience.2., namely query and reporting mechanism, further refined as resource adaptation mechanism. In the inquiry mechanism, the resource manager ask Forwarding component, global network state information static, the mechanism of real-time, but will bring extra control overhead is greater. In reporting mechanism, local network state information active push forward component dynamic, the mechanism is almost no extra control overhead, but the real time affected by network traffic. Then, in the prototype system realized two kinds of mechanism, design different traffic scenarios, real-time verification and comparison of two kinds of mechanism. Finally, using real traffic data, in two aspects of cost and real-time balancing mechanism of two kinds of compromise, the actual application effect were analyzed and evaluated. The results show that the real-time reporting mechanism is not asked mechanism however, report mechanism has no extra control overhead advantage, although there are some delay and error, but the degree in the acceptable range, therefore, the reporting mechanism to transfer information is perceived The.3. presents a wisdom path allocation method based on traffic awareness, to support collaborative intelligence network "dynamic resource scheduling". Firstly, the real-time traffic matrix calculated as a constraint on the load balancing routing optimization model, using the Lagrange dual method, the original problem into a dual problem. Then, to seek the optimal solution of dual problem, design a load balancing algorithm based on flow path allocation matrix. Finally, through the simulation experiment and prototype system test, verify the proposed load balancing algorithm and additional control overhead. The results show that this method can effectively avoid congestion, network load balancing, and the additional control overhead are within the acceptable range of.4. presents a cooperative cache optimization method and the energy consumption of the network, to support collaborative intelligence network "choose dormancy" The function of wisdom. Cooperative forwarding components in the network not only perform simple forwarding, also has the cache and sleep function. Therefore, in order to take into account the cache performance at the same time, the maximum saving network energy consumption, establishes a model of multi-objective mixed integer linear programming problem. Then, the cache performance optimization objectives into constraints, so much the target problem is transformed into a single objective problem, design and component selection algorithm to solve the dormancy. Finally, simulation of the real network traffic, respectively in the three real network topology, a comparative assessment on the energy saving effect of the proposed algorithm and the existing algorithms. The results show that the proposed algorithm of the energy saving effect is better than the other algorithms; redundant link more and better network, energy saving algorithm; in the implementation of the method of idle network, the remaining network still can keep load balance well.

【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

【相似文献】