网络资源协同管理与多目标虚拟网络映射问题研究

Network virtualization is considered a promising solution to the gradual ossification of the Internet. In a network virtualization environment, software and hardware of many physical networks, which may be owned by different infrastructures, are cooperatively managed in a uniform method and shared by many virtual networks and applications which are often isolated from others. Our study will be conducted in the following way. First, an autonomous system based management architecture of network resources from different infrastructures and heterogeneous substrate networks is proposed and implemented. Therefore, network resources are maintained and distributed on various hierarchical agents. Second, multi-objective virtual network mapping problem (MOVNMP) is formulated, which is NP-hard. After MOVNMP is theoretically analyzed, multi-objective algorithm based on swarm intelligence is proposed to attack it. At last, considering dynamic requirements of resources in virtual networks, a dynamic allocation model and the corresponding solution are designed by employing dynamic optimization methods so as to maintain load balance and maximize overall utility. Our study will benefit the Internet a lot by overcoming its ossification.

网络虚拟化是克服当前因特网僵化问题的一种重要技术。网络虚拟化技术将网络的软硬件资源进行整合，使得能在一个公共的底层物理网络上同时运行多个网络架构或网络应用。围绕网络资源的管理与分配，本研究提出设计一种基于自治域的网络资源协同管理架构，实现对隶属于不同基础设施提供商、异构网络资源的分布式、分层次管理与使用；针对虚拟网络映射问题，建立虚拟网络映射的多目标数学模型，并对该NP-Hard问题进行理论分析，设计基于群体智能的启发式求解算法；针对虚拟网络资源的动态分配问题，建立基于负载均衡和效用最大的网络资源分配模型，借助动态优化方法，设计相应网络资源动态分配方案。通过对以上几个网络虚拟化核心技术的研究，以期促进因特网僵化问题的解决。

本项目的研究工作包括：（1）研究了协同进化生物地理学优化算法、多目标优化理论及群搜索算法理论与方法，建立了虚拟网络映射的数学模型，设计实现了基于微粒群算法和群搜索算法的求解方法；（2）研究绿色节能虚拟网络映射模型和可生存性虚拟网络映射模型，设计提出了基于群搜索算法的求解方法和支持路径切割和资源共享的可生存性虚拟网络映射求解算法；（3）基于用户资源需求动态变化的实际情况，研究了虚拟网络资源的动态分配问题，设计实现了基于负载均衡和SVM、神经网络的动态资源预测和分配算法，实现网络资源动态分配；（4）设计实现了一种基于自治域的网络资源协同管理与优化仿真平台（NRCMOP），实现对隶属于不同基础设施提供商、异构网络资源的分布式、分层次管理与使用的仿真。（5）在《Applied Intelligence》、《International Journal of Web and Grid Services》、《Multimedia Tools and Applications》、《Journal of Internet Technology》等SCI、EI期刊发表论文12篇，在EI收录国内外学术会议论文集发表论文24篇，在中文核心期刊发表论文6篇，授权国家发明专利4项，实审中专利2项，共有15名博士、硕士研究生参与项目研发。