Dynamic predictive weighted fair queueing for differentiated services

The main objective of this paper is to establish that it is not efficient to use weighted fair queueing (WFQ) with a static allocation of bandwidth for the different traffic classes in a differentiated services environment and to propose an extension to WFQ, based on prediction techniques of realistic traffic models, in order to provide it with the necessary adaptability to cope with the dynamic nature of Internet traffic. We call our proposed scheme dynamic predictive weighted fair queueing (DP-WFQ). Its performance is evaluated via simulations. In addition to WFQ, our proposal is also compared to other algorithms, such as WF2Q and LTO-WFQ. Our results show the superior performance of the new algorithm in terms of the number of transmitted packets, buffer overflows, and delay