基于Q-learning的跨域服务链映射机制

摘要

A virtual service resources controlling architecture with regional centralized management and global coordinated scheduling was proposed for the problem of cross-domain service chain mapping in SDNFV environment. On this basis, an effective mapping framework was built and the cross-domain mapping problem was modeled as an ILP with the purpose of minimizing mapping cost. A partitioning algorithm was designed to solve the problem based on Q-learning mechanism under this framework. Simulation results show that the performances of this method are better than other traditional methods on average partition time, average mapping cost, and acceptance ratioof service chain mapping request.%针对软件定义网络功能虚拟化环境下跨域服务链映射问题,提出了一种区域集中管理、全局协同调度的虚拟服务资源管控架构.在此基础上,建立了一种有效的跨域服务链映射框架,在此框架下将跨域服务链映射问题建模为以最小化映射开销为目标的整数线性规划问题,并基于Q-learning机制设计跨域服务链构建请求分割算法进行优化求解. 仿真实验表明该方法在平均分割时间、平均映射开销和服务链构建请求接受率等方面相较传统方法具有更优的表现.