On the use of packet classes in communication networks to enhancecongestion pricing based on marks

This note explores the use of packet classes to enhance Internet congestion pricing based on marks in the framework of Kelly et al. (1997). Two packet classes are used in order to exploit differences in the multidimensional quality-of-service (QoS) requirements within the population of users. Simple scenarios for a link with a finite buffer with and without burstiness are investigated. It is concluded that the use of packet classes provides a margin for error in provisioning and operating a packet network with congestion pricing