Service oriented and fine-grained bandwidth allocation in large scale IP networks

The diverse types of service beared on the large-scale IP network makes bandwidth allocation a research focus. This paper proposes a dynamic and fine-grained bandwidth allocation method which is suitable for large-scale network. We extends Norros' Fractional Brownian Motion (FBM) model to make it fit for the allocation of service bandwidth in IP network. Through identifying and analyzing different services, we can get the service characteristic parameters and achieve the demanding bandwidth of different services. Then, we optimize scheduling algorithm of Diffserv scheme by allocating queue polling weights dynamically according to the weights of demanding bandwidth of different services. We build up a simulation network which nodes employ Diffserv scheme and deploy our optimized schedule algorithm. The simulation results show that the QoS of different services can be assured effectively.