On providing bandwidth and delay guarantees using the revenue criterion based adaptive WFQ

Abstract

This paper proposes the adaptive resource allocation model that is based on the WFQ service policy. This model ensures QoS requirements and at the same time tries to maximize a service provider's revenue by manipulating weights of the WFQ policy. The model is flexible in that different network services are grouped into service classes and are given such QoS characteristics as bandwidth and delay. To adjust dynamically weights it is proposed to use the usage-based revenue criterion that enables the allocation of free resource between service classes. The simulation considers a single node with the implemented model that serves several service classes with different QoS requirements and traffic characteristics. It is shown that the total revenue can be increased due to the allocation of unused resources to more expensive service classes. Though the simulation has revealed that certain QoS threshold are violated, the percentage of them is insignificant and can be easily tolerated by many network applications.