Performance of packet scheduling methods with different degree of fairness in HSDPA

The present paper focuses on assessing the performance of the packet scheduler in HSDPA. The investigation has especially concentrated on assessing different scheduling methods with different degrees of fairness for non-delay-sensitive elastic traffic. The fairness ultimately determines the QoS perceived by the users. The results show that very unfair scheduling methods provide the highest HSDPA cell throughput for best effort NRT traffic, although, under heavy load situations, the unfairness of these policies causes the starvation of users under poor average radio propagation conditions. Under minimum user throughput guarantees, the proportional fair algorithm provides the highest cell capacity of all tested algorithms. With this scheduler, the provision of user throughput guarantees of CA, 128 and 384 kbit/s (for at least 95% of the users) incurs on a cell throughput reduction respectively of around 10, 30 and 80% relative to the maximum cell throughput for best effort traffic (i.e. no user throughput guarantees at all).