Analysis of sfqCoDel for Active Queue Management

Abstract

The availability of cheaper and high capacity Random Access Memory (RAM) has resulted in the growth of buffer size in all the computing devices. This aberrant increase of buffer capacity in network devices has resulted into high latency, leading to reduced throughput; thus decreasing the tendency of absorbing spontaneous burst of traffic. The requirement for Active Queue Management (AQM) has been evident for decades. These solutions require various parameter configuration and are dependent on a particular network condition to work efficiently. Hence an algorithm which is simple, efficient, does not require setting of parameters and works seamlessly irrespective of the network condition is required. Even though Controlled Delay (CoDel) is parameterless and adapts to dynamically changing link rates with no negative impact on utilization, it deviates from its primary purpose of reducing congestion when there is an increase in RTT and when congestion level varies abruptly. As a consequence, a variant of CoDel called Stochastic Fair Queue CoDel (sfqCoDel) is simulated and compared. The Stochastic Fair Queue CoDel proactively drops packets which occupy reasonably larger bandwidth as compared to CoDel, which proactively drops packet irrespective of the bandwidth consumption by packets. This paper aims to perform a comprehensive analysis of Stochastic Fair Queue CoDel for Active Queue Management. A comparison is also carried out between sfqCoDel with CoDel. The sfqCoDel appears to be much better than CoDel in certain areas where CoDel fails to perform well.