MDP-based dynamic update of aggregated information for hierarchical QoS routing

Abstract

ATM private network-to-network interface (PNNI) adopts hierarchical routing to achieve scalability in large ATM networks. For performing hierarchical routing efficiently, the network consists of peer groups, and peer groups advertise the aggregated information periodically which is based on a time-based update interval. The PNNI time-based update policy is inadequate to cope with dynamic network traffic. Therefore, in this paper, we propose two dynamic update policies, referred to as the dynamic cost-based update (DCU) policy and the DCU with hysteresis (DCUH) policy, to improve the accuracy of aggregated information and the performance of hierarchical routing, while decreasing the frequency of re-aggregation and distribution. Furthermore, in the past, the definition of cost function has received much less attention. In this paper, we study the Markov decision process (MDP) theory to define the cost function. Based on the MDP approach, we compare the DCU policy with PNNI time-based update (PNNIU) policy, full update (FU) policy, and logarithm of residual bandwidth update (LRBU) policy. Our simulation results indicate that the proposed DCU and DCUH policies yield better performance while significantly reducing the frequency of re-aggregation and the amount of distributed aggregation information. The overhead of re-aggregation of DCUH is lower than that of DCU due to the effect of hysteresis.