Network capacity allocation for traffic with time priorities

Abstract

The packet switching techniques are undergoing evolution. The conventional 'best effort' approach will no longer be the dominant service. The next generation of IP networks must provide QoS to customers. Inadequacy is obvious when the conventional capacity allocation (CA) models are applied to the new IP architecture. In this paper, we propose several CA models that characterize: (1) the service priority scheme; (2) the service preemption scheme; and (3) the non-Poisson traffic in which the packets follow heavy tailed distributions.