Quality of service management in IP networks

Abstract

Absfracf-The DiffServ architecture provides a scalable mechanism for QoS introduction in an IP network. The idea o f DifServ is based on the aggregation of traffic flows at an ingress (or egress) point of a network and IP packet marking for different priority flows, according to several classification criteria. In this paper the problem of the improvement and fairness of absolute QoS provisioning to paths established along a DiffServ network on a per router basis is considered. The service rate reconfiguration problem of a router's output link is formally defined, mathematically formulated and solved by means of efficient heuristic algorithms, providing good solutions in reasonable time. Finally, an indicative set of results is provided and concluding remarks are made.Keywords-Quality of Service, Di/fprenfinfed Services, Service Rare Reconfgurotion. 0-1 Lineor Programming. 1. INTRODUCTlONService differentiation is considered to be of outmost importance for QoS provisioning in IP networks, due to the high variations of the connection requirements posed by Internet users and the statistical in general nature of the gencrated traffic, which the last years is presenting an exponential increase. The research community has concentrated on two different techniques to provide QoS differentiation to customers of packet switched networks. First, the Integrated Services (IntServ) [1][2] and the Differentiated Services (DiffServ) [3][4][5] approach. The major difference between IntServ and Diffscrv architecture is the granularity of service differentiation. The IntServ concept lies in resource reservation notion, while in DiffServ model IP traffic is classified into finite, predefined service classes (on the basis of the demand requirements and characteristics) that receive different routing treatment. DiffServ achieves scalability and manageability by providing quality per traffic aggregate and not per application flow, while on the other hand IntServ approach faces potential problems.Two directions exist in the Diffjerv architecture, the relative and the absolute. In absolute DiffServ [6] architecture, strict QoS parameters are defined for each service class. An admission control scheme is used [7] t6 provide QoS guarantees as absolute bounds of specific QoS parameters such as bandwidth, packet transfer delay, packet loss rate, or packet delay variation (jitter). For any accepted user the appropriate resources are reserved and the level of performance of his connection is assured. The relative DiffServ model [B] provides QoS guarantees per class in reference to guarantees given to other classes. The only assurance from the network is that higher classes receive better service treatment than lower classes. Proposals for relative per class DiffServ QoS define service differentiation qualitatively [9][10] in terms that higher classes receive lower delays and losses from lower classes. Relative service differentiation is a simple and easy deployed approach compared to the absolute service differentiation.Taking into account t...

Proportional delay differentiation provision by bandwidth adaptation of class‐based queue scheduling

“…* Case 2 (Prediction of ADC violation): In the event that ADC violation is predicted, the algorithm relaxes RDCs and tries only to satisfy the strict ADCs. To this respect, our model adopts the algorithm described in Reference [25]. Specifically, a portion of the service rates of lower priority classes that meet their ADCs are exploited in order to increase accordingly the rate of the ADC violated classes.…”