Regresar

Evaluation of Utility Function Algorithm for Congestion Control in Computer Networks

Abstract:

In computer networks, the huge amount of heterogeneous transmitted data causes the performance of the networks to decrease. Congestion is a problem that causes network underutilization and packet loss. Two approaches have been identified in the literature to treat congestion: mitigation, mainly with the management of window size; and, avoidance, with an explicit assignment of transmission rates. In this paper, congestion control using avoidance is proposed with the implementation of a Utility Function Algorithm (UFA). UFA is a variation of the centralized B-Neck algorithm that uses the max-min fair criterion, UFA integrates the quality of service (QoS) approach. UFA assigns each session the bandwidth it needs according to its type of traffic and considers the traffic of all sessions to distribute the available bandwidth so that all sessions can have a better level of QoS. In the experimentation, three test scenarios with variations of mixed traffic and different link capacities are specified. B-Neck and UFA were implemented in Matlab to measure their performance using metrics: the transmission rate allocation, utility of each session, and the Gini coefficient to measure fairness in allocation. The results show that UFA achieves better performance in all scenarios because it allocates only the bandwidth that each session needs.