Markov Chains Based Dynamic Bandwidth Allocation in DiffServ Network

This letter proposes a Markov chain based model for dynamic bandwidth allocation in DiffServ networks. At a time-slot, the proposed Markov chain is used to predict the bandwidth requirement at the next time-slot, and resource is then allocated accordingly. Such a pre-allocation scheme can effectively reduce the operation overhead in bandwidth allocation and further reduce the connection blocking probability. We present numerical results showing that our dynamic bandwidth allocation mechanism can reduce the network blocking probability by one order of magnitude, compared with the existing bandwidth borrowing mechanism.