Transient performance analysis of the selective drop buffer acceptance scheme with responsive traffic

This paper considers the selective drop (SD) algorithm, a packet aware buffer acceptance algorithm used with the unspecified bit rate ATM service category. The influence of the parameters of the SD algorithm on the transient performance results (efficiency and fairness) is studied when starting from an unfair start situation. This is done using an analytical model where two responsive sources send their traffic in packets consisting of cells via a buffer on which the SD algorithm is implemented. Three different scheduling algorithms are used in combination with the SD algorithm. Based on an extensive set of numerical examples, observations are made. From these results, one of the most important conclusions is that a large improvement of the fairness results is seen when SD is implemented, irrespective of the exact setting of its parameters. On the efficiency results, these parameters have however a more important effect. So it is recommended to implement SD to increase the fairness, but with a parameter setting focussed on the efficiency results