Controlling queue size in a single server system

The authors revisit the well-known problem of active queue management in queueing systems and show numerically that the efficiency of the algorithms based on dropping functions can be improved. Three different approaches are discussed: randomization of the parameters of RED-type dropping functions, use of the arbitrary dropping functions either without or with parallel tracking of the remaining/elapsed service time. The relative efficiency of these three approaches is demonstrated on $M/D/1/N$ queue with implemented RED scheme. For various initial settings, simulation results are given, which show that the new active queue management algorithms outperform RED schemes and the gain increases with the increase of the system load.