一种基于流量预测的资源动态管理算法

Dynamic resource management algorithm based on Internet traffic prediction

分析了基于有效带宽理论(effectivebandwidththeory)的静态资源分配方案的弊端。在此基础上,提出了一种基于流量预测的资源动态管理算法,并且把资源动态管理算法具体应用到QoS的区分服务(DifferentiatedService)体系中,算法在区分服务网络的边缘路由器上实现。最后,在ns 2的仿真环境下对两种算法进行了比较,试验结果证明无论在丢包率还是链路利用率上,资源动态管理算法都比静态资源分配方案有明显的优势。

The static resource allocation algorithm based on the theory of effective bandwidth was introduced firstly. When the real behavior of Internet traffic is taken into account, this algorithm is inefficient. So here a dynamic resource management algorithm based on Internet traffic prediction was proposed to take the place of it. This algorithm was applied to a Differentiated Service network, and implemented on the boundary node. The basic idea under this algorithm was to allocate resources (bandwidth/buffer size) between different kinds of flows dynamically, according to the result of prediction. At last, ns-2 was used to run the simulation and find out the lost packets rate and output link utilization of this algorithm, which were superior to those of the static resource allocation algorithm.