Enhancing Congestion Control in TCP Networks: Geometrical and Physical Interpretations of Queue Dynamics for Random Early Detection

TCP Congestion control using Active Queue Management (AQM) techniques, especially Random Early Detection (RED), is an evolving and persisting difficulty due to the dynamic nature of network traffic. Achieving a unique parameter configuration that suits all traffic natures might be challenging, and erroneous settings could dramatically reduce the overall network performance. Recent advancements in data transportation proposed optimal parameter configuration to minimize the parameter sensitivity of RED and maintain high performance. Random Early Detection (RED) strategy and its variants use an average queue size parameter to decide on the optimal drop rate of packets to absorb congestion. In this paper, we highlight some of the drawbacks of RED and propose two methods for drop rate calculation depending on queue behaviors to overcome RED's disadvantages. We suggest a queue-trend line, an increasing or decreasing line that makes an angle with the timeline. In addition, we present the ideas of geometrical and physical interpretations of queue dynamics. First, the geometrical method assumes that the queue-trend line is the tangent of the queue curve. Accordingly, it can be interpreted as a congestion severity metric, leading to the development of the Queue Dynamics Coefficient (QDC) parameter, which can be utilized to tune packet drop rates. Second, the physical method considers a line between two points in a queue curve as the queue-trend line. Based on that, the slope of the queue-trend line is interpreted as a congestion indicator. A triangular shape formed by the physical method led to the concept of the average speed of queue dynamics, which is beneficial to estimating a tuned response for congestion. Moreover, this paper presents several open issues in TCP congestion control to emphasize the need for ongoing research. Overall, this article proposes promising methods to enhance the performance of RED-based strategies.