Active Queue Management for Congestion Control: Performance Evaluation, New Approach, and Comparative Study

Abstract

Network congestion is one of the most important problems that effect Quality of Service (QoS). Several Active Queue Management (AQM) algorithms have been developed to avoid congestion problem by controlling the queue length in routers. However, an important problem arising with the current AQM algorithms is that most of the current algorithms handle different traffics by the same strategy. This problem may lead to performance degradation especially for real-time applications ssuch as video and audio traffics. This paper first presents a performance evaluation of the current AQM algorithms. It then presents a new AQM algorithm, called Dynamic Queue RED (DQRED), to guarantee efficient QoS to both real-time and non-real-time traffics. Finally, a comparative study is done between the proposed DQRED algorithm and the most recent AQM algorithms by using the network simulator (NS-2) considering different QoS metrics.

A Fully Adaptive Active Queue Management Method for Congestion Prevention at the Router Buffer

“…The results are illustrated in Figs. [16][17][18][19][20]. The results concur with the previous results, with stability shown for AAQM and RED; however, AAQM achieved the best results regarding throughput and delay with a low dropping rate.…”

“…Multiple Average-Multiple Threshold (MAMT) [16] extended RED by implementing different dropping techniques for different traffic types, as each requires different qualities, such as low delay, low loss, or maximum utilization. Dynamic Queue RED (DQRED) [17] and Probability RED (PRED) [18] use various queues, each for a different traffic type, to achieve high quality of service (QoS). Predictive Hebb AQM (PHAQM) [19] was developed based on model predictive control (MPC) and optimized using Hebb learning rules from the neural network control theory with a delay indicator.…”

“…Unlike the existing AQM methods, the proposed method adaptively updates the value of the Dp and utilizes all traffic and router evidence. The non-adaptive methods, including CHOKe [14], XCHOKe [15], MAMT [16], DQRED [17], PRED [18], NVARED [20], NRED [21], QueuedRED [22], RaQ [23], Effective RED (ERED) [24], VSC [25] NPD [26], AQMRD [27], and S-RED [28] rely on non-adaptive queue length information, which may influence the stability and performance of temporary traffic changes. The PHAQM [19] method solely utilizes delay as an indicator in nonadaptive mechanism.…”