Dynamic queue management using neural network based on balanced RED
Mona Amoli Diva, Mohammad Teshnehleb · 2014
In this paper we present a new technique for network congestion avoidance and control, based on early Balanced Random Early Detection algorithm (BRED). We have optimized BRED algorithm so that the new algorithm can dynamically detect non-adaptive flows and limit receive rate from them, to provide fairness between flows and avoid occurring congestion and buffer overflow. In this method we have used time delay line neural network as system's core to detect and separate adaptive and non-adaptive flows. We will discuss about the algorithm and compare simulation results with BRED and Drop Tail.