时间和能量感知的贝叶斯虚拟网映射

Time and energy aware virtual network embedding using Bayesian theory analysis

针对虚拟网的节能映射问题,建立了结合时间和能量感知的虚拟网映射算法。在对节点和路径的评价标准中加入了时间因素,综合考虑了物理资源的运行时间等因素,用概率理论辅助分析了每个虚拟节点的多个可用物理节点被选中的概率。在节点选择阶段,综合考虑底层节点的剩余资源量、CPU资源利用率增量、节点开启情况和是否延长使用时间等因素,并使用条件概率理论辅助分析得到各可用节点的重要性;在链路选择阶段,综合考虑链路开启情况、延长使用时间和链路长度等因素。不仅使虚拟网请求映射在当前较小的节点和链路集合中,而且映射到了延长时间较短的设备上。实验结果表明,与未考虑时间因素的方法相比,该方法能带来更好的性能和更低的能耗。

Aiming at the energy consumption problem in virtual network embedding, a virtual-network-embedding algorithm was proposed by combining the time and energy aware. Taking the running time during the evaluation of physical nodes and physical paths into account, it considered multiple factors which included the processing time of physical devices, and used probability theory to help analyze the selected probability of each available physical node for a virtual node. During the selection of substrate nodes, the factors of remaining resources, the increment of CPU utilization, the switch state and the amount of extended time of physical nodes were considered. The theory of conditional probability was further used to analyze the importance of available nodes. The factors of the switch state, the amount of extended time and the length of physical paths were also considered. The proposed approach could effectively map the current virtual network request onto a smaller set of nodes and links which are switched on, and also the devices which have less amount of extended time. Experimental results show that the proposed approach has better performance, and can effectively decrease energy consumption comparing with the methods without taking the time factor into consideration.