The Good, the Bad and the WiFi: Modern AQMs in a residential setting

Abstract

a b s t r a c tSeveral new active queue management (AQM) and hybrid AQM/fairness queueing algorithms have been proposed recently. They seek to ensure low queueing delay and high network goodput without requiring parameter tuning of the algorithms themselves. However, extensive experimental evaluations of these algorithms are still lacking. This paper evaluates a selection of bottleneck queue management schemes in a test-bed representative of residential Internet connections of both symmetrical and asymmetrical bandwidths as well as WiFi. Latency under load and the performance of VoIP and web traffic patterns are evaluated under steady state conditions. Furthermore, the impact of the algorithms on fairness between TCP flows with different RTTs, and also the transient behaviour of the algorithms at flow startup is examined. The results show that while the AQM algorithms can significantly improve steady state performance, they exacerbate TCP flow unfairness. In addition, the evaluated AQMs severely struggle to quickly control queueing latency at flow startup, which can lead to large latency spikes that hurt the perceived performance. The fairness queueing algorithms almost completely alleviate the algorithm performance problems, providing the best balance of low latency and high throughput in the tested scenarios. However, on WiFi the performance of all the tested algorithms is hampered by large amounts of queueing in lower layers of the network stack inducing significant latency outside of the algorithms' control.

Flent

“…Figure 4 shows a CDF of the latency measurements across all test repetitions for the CoDel and PIE AQMs for both CUBIC and BBR. Consistent with earlier evaluations of the AQMs [10], CoDel achieves lower delay than PIE, since it tends to drop more aggressively. However, because this di erence is more pronounced when using CUBIC (the leftmost and rightmost lines), it means that the relative performance of BBR and CUBIC is reversed depending on the AQM: With CoDel, the queueing delay is lower when using CUBIC that when using BBR, while the reverse is true for PIE.…”

“…For our testbed evaluation, we re-use the testbed from a previous study of AQM algorithms [10]. The testbed consists of ve nodes, connected as depicted in Figure 1.…”

Measuring Latency Variation in the Internet

“…However, experience from more controlled environments (such as experiments performed to evaluate AQM algorithms [14]) suggests that queueing delay can be a significant source. Due to the magnitude of the variation we see here, we conjecture that this is also the case in this dataset.…”

Piece of CAKE: A Comprehensive Queue Management Solution for Home Gateways

“…We use the Flent testing tool [23] to run the tests, and the data files are available on the companion web site. 1 Unless otherwise stated below, all tests are run on a symmetrical 10 Mbps link with 50 ms baseline latency. Our basic test is the Real-Time Response Under Load test, which consists of running four TCP flows in each traffic direction, along with three different latency measurement flows [24].…”