自适应路由服务合成:模型及优化
发布时间：2019-01-04 12:38
【摘要】：当前,新型网络应用不断涌现,用户对不同类型应用的通信需求也呈现出多样化和个性化的特点.面向用户频繁产生和变化的通信需求,网络服务提供商(Internet service provider,简称ISP)通常以不断地购买及部署大量新型的专用网络设备的方式来应对,导致其运营成本高昂,资源浪费严重,网络建设与发展的可持续性差.对此,从软件角度出发,考虑路由功能重用,通过选择合适的路由功能,在通信路径上为应用合成定制化的路由服务,满足用户差异化的需求.基于网络功能虚拟化(network function virtualization,简称NFV)和软件定义网络(software-defined networking,简称SDN),提出了一种自适应路由服务合成机制,运用软件产品线技术构建路由服务产品线,作为路由功能选择和路由服务优化的基础.基于机器学习,运用多层前馈神经网构建路由服务离线模式和在线模式两阶段学习模型,对路由功能选择及组合进行持续学习和优化,实现路由服务的定制化目标,以提高用户的服务体验.进行了仿真实现,研究结果表明,所提出的模型是可行和有效的.
[Abstract]:At present, new network applications are emerging, and users' communication needs for different types of applications are diversified and individualized. Facing the frequent and changing communication demand of users, (Internet service provider, (ISP), a network service provider, usually deals with it by constantly purchasing and deploying a large number of new types of special network equipment, which results in high operating cost. The waste of resources is serious and the sustainability of network construction and development is poor. From the point of view of software, considering the reuse of routing function and selecting the appropriate routing function, the application composition customized routing service can be used in the communication path to meet the needs of user differentiation. Based on network function virtualization (network function virtualization,) and software defined network (SDN), an adaptive routing service composition mechanism is proposed, which uses software product line technology to construct routing service product line. As the basis of routing function selection and routing service optimization. Based on machine learning, a two-stage learning model of off-line and on-line mode of routing service is constructed by using multi-layer feedforward neural network. The continuous learning and optimization of routing function selection and composition are carried out to achieve the customized goal of routing service. To improve the user's service experience. The simulation results show that the proposed model is feasible and effective.
【作者单位】： 东北大学软件学院;东北大学信息科学与工程学院;
【基金】：国家自然科学基金(61572123) 国家杰出青年科学基金(71325002)~~
【分类号】：TP393.09
本文编号：2400303
[Abstract]:At present, new network applications are emerging, and users' communication needs for different types of applications are diversified and individualized. Facing the frequent and changing communication demand of users, (Internet service provider, (ISP), a network service provider, usually deals with it by constantly purchasing and deploying a large number of new types of special network equipment, which results in high operating cost. The waste of resources is serious and the sustainability of network construction and development is poor. From the point of view of software, considering the reuse of routing function and selecting the appropriate routing function, the application composition customized routing service can be used in the communication path to meet the needs of user differentiation. Based on network function virtualization (network function virtualization,) and software defined network (SDN), an adaptive routing service composition mechanism is proposed, which uses software product line technology to construct routing service product line. As the basis of routing function selection and routing service optimization. Based on machine learning, a two-stage learning model of off-line and on-line mode of routing service is constructed by using multi-layer feedforward neural network. The continuous learning and optimization of routing function selection and composition are carried out to achieve the customized goal of routing service. To improve the user's service experience. The simulation results show that the proposed model is feasible and effective.
【作者单位】： 东北大学软件学院;东北大学信息科学与工程学院;
【基金】：国家自然科学基金(61572123) 国家杰出青年科学基金(71325002)~~
【分类号】：TP393.09
【相似文献】
相关硕士学位论文 前1条
1 陈霄;一种可变SaaS应用的开发方法研究[D];复旦大学;2013年
，本文编号：2400303
本文链接：https://www.wllwen.com/guanlilunwen/ydhl/2400303.html
