Maximizing EF-class traffic in DiffServ Internet

DiffServ is becoming the primary QoS architecture for the Internet. The DiffServ architecture classifies traffic into different categories. Among them Expedited Forwarding (EF) is the premium class and has the highest priority. Major applications of the EF-class include real-time services like voice-over-IP and services that can compete with circuit-switched leased lines. Maintaining the QoS of the EF-class traffic is a challenging problem as service-level-agreements (SLA) in the DiffServ Internet provide no or only partial destination information. This problem is tightly coupled with QoS routing in these networks. In this paper we present an analytical framework for computing and maximizing the amount of admissible EF-class traffic in the DiffServ Internet and DiffServ/MPLS networks. We present two general approaches and compare their performance in terms of the amount of admissible EF-class traffic. We also show how traffic engineering of a MPLS network can further increase the amount of admissible EF-class traffic in a DiffServ/MPLS network.