Maximizing network resource utilization through dynamic delay allocation adjustment

Abstract

Quality of Service (QoS) has been an important topic in network research, and many solutions have been proposed to address QoS related issues. In this project, we focus on the delay requirement partition issue for maximizing the network utilization. We first survey conventional schemes, and point out their limitations: they all perform static allocations based on instant load situations, which may cause imbalanced reservations and bottleneck links. We then propose a novel Dynamic Allocation Adjustment (DAA) algorithm to address these problems by dynamically adjusting the existing reservations for earlier admitted flows. DAA not only spreads traffic evenly onto intermediate links but also balances link loads in a broader range. As a result, link congestion on a flow path is alleviated and its total reservation is reduced. On the other hand, DAA addresses the bottleneck link problem. We conducted our simulations on both of symmetric and asymmetric topologies, with uniformly distributed traffic and several types of imbalanced traffic. The results show that the improvement of system utilization is over 30%.