Dynamic queue velocity-based probability function for congestion avoidance in high-speed networks

Abstract

Network congestion has serious negative impact on the utilization and efficiency of the network resources and can result in network collapse. Minimizing congestion occurrence in a network would optimize the network performance and satisfy end users requirements. As high-speed networks are designed to offer widespread services, including real time multimedia, an efficient congestion avoidance scheme is required to keep the network functioning within its capacity. This paper introduces a queue velocity-based drop probability function that is essential for any proactive queue management mechanism. It is a dynamic function of the instant (current) queue size that enables a queue management mechanism to provide more specific information regarding condition of the router buffer upon packet arrivals. It allows the router to differentiate a range of congestion conditions and to treat them accordingly. It can help lowering the chance of unnecessary packet drops, prevents excessive oscillations in the current queue length, and improves the link utilization.