An Improved RED Algorithm with Input Sensitivity

Random Early Detection (RED) is a recommended scheme of Active Queue Management (AQM) to avoid network congestion by Internet Engineering Task Force (IETF). RED being sensitive to its control parameters setting and traffic load behaves differently during different phases of scenarios and its performance degrades. To solve the above problems, we have proposed an improved algorithm LTRED, L stands for queue length and T stands for threshold, which incorporates the impact of load variation in early congestion notification along with tuning of threshold parameter of RED. Our approach is validated by doing extensive simulations on ns-2 (network simulator) and comparisons are done with RED, ARED, and AVQ well-known AQMs. Our approach has shown its superiority in terms of packet loss, early congestion indication, better link utilization, and improved goodput. In this work, mismatch between macroscopic and microscopic behavior of queue length of RED is also addressed and reduction in mismatch is achieved. In addition, LTRED makes very few changes to the original RED without making the system complex.