Performance behavior evaluation of Internet congestion control policing mechanisms

Performance behavior is an important issue in the design and implementation of an efficient Internet congestion control policing mechanism. The effectiveness of such a mechanism can be measured by packet loss probability, bandwidth allocation, packet delay, throughput or other quality of service measures. In this paper, we carry out a comprehensive study to investigate the performance behavior of four selected policing mechanisms for the Internet namely: token bucket (TB), jumping window (JW), triggered jumping window (TJW) and exponentially weighted moving average (EWMA). Three types of bursty sources modeled as On/Off, Poisson and batched Poisson processes are utilized. Three criteria are used to evaluate the performance behavior of the selected policing mechanisms. These are the average packet delay, the average packet loss probability and the average number of lost credits. Computer simulations were used to arrive at various conclusions regarding the dependence of performance on source traffic characteristics and policing mechanism parameters. Furthermore, a comparison of the performance behavior of the selected policing mechanisms was carried for different input traffic characteristics.