An Extended GI~X/M/1/N Queueing Model for Evaluating the Performance of AQM Algorithms with Aggregate Traffic

摘要

TCP/AQM models focus on responsive long-lived TCP flows and are often used as a guideline to design new AQM algorithms. Nevertheless, the Internet traffic is aggregate. Besides responsive long-lived TCP flows, there are 70%-80% unresponsive short-lived TCP flows and UDP flows which are ignored by these TCP/AQM models. In this paper, we first extend the GI~X/M/1/N queueing model with batch arrivals by thinning of input flows and then use probability distributions of aggregate traffic as input to this extended model to evaluate and compare the performance of four classical AQM algorithms: TD, RED, GRED and Adaptive RED.