Which AQM fits IoT better?

Abstract

Most of the Internet traffic is carried by Transmission Control Protocol (TCP) nowadays, even in the case of real-time services and IoT environments. A key point of each TCP variant that defines the specific variant trademark is the congestion control algorithm. This congestion control, alone, is not able to mitigate the congestion problem completely; the most efficient solution, to be coupled with it, is the Active Queue Manager (AQM). In this paper, we analyze the response to congestion provided by the novel AQMs when coupled with popular TCP variants and, has happened in IoT networks, with a mix of TCP variants simultaneously active on the network. We provide extensive simulations within ns-3 to highlight what is the better AQM solution for IoT networks by considering the TCP algorithm and other environmental conditions like the number of active nodes, the network RTT and the presence of multiple RTT flows. We collected network metrics like throughput, goodput, latency, RTT variation, flows' fairness and a simple drop analysis.

Nesnelerin Kuyruk Yönetimi: Nesnelerin İnternetine Özel Kuyruk Yönetim Yaklaşımı

“…Grazia et al (Grazia et al 2017) provides a simulation-based comparison of the selected ones of these algorithms in terms of goodput, throughput, RTT variation, fairness, etc. in order to study suitability of these algorithms for an IoT environment.…”

“…ARED shows up with the best performance in a highly congested environment, however it diminishes its performance when it encounters with an environment with multiple RTT flows. It is also good at fine-tuning the trade-off between goodput and delay performance (Grazia et al 2017). However still as it does not value the content and network exploiting nodes, and since sender nodes do not cooperate as they do in AQMoT approach, it may not be fully "adaptive" to IoT environment.…”

“…Moreover, this algorithm proposes a good fairness index between flows (Feng et al 2002). According to (Grazia et al 2017), PIE, which also overcomes GREEN, has a resistance to all the stress factors, and it has one of the best trade-offs between channel exploitation and delay. Coping with channel exploitation is actually a big part of coping with the whole IoT network.…”

“…This algorithm uses different approach of drop probability calculation by not only using the current estimation of the queuing delay, but also sensing the direction of where the delay is shifting, which is not a simple random drop possibility. Grazia et al (Grazia et al 2017) provides a simulation-based comparison of the selected ones of these algorithms in terms of goodput, throughput, RTT variation, fairness, etc. in order to study suitability of these algorithms for an IoT environment.…”

Supervised Learning of Neural Networks for Active Queue Management in the Internet

“…In this context, the information obtained from a classic PI controller is used as a packet dropping function [ 16 , 17 , 18 ]. The article [ 19 ] highlights the advantages of the PIE (Proportional Integral Enhanced Controller) algorithm. The authors state that mechanism easily adapts to varying transmission conditions and turned out to be a compromise between the degree of queue utilization and transmission delays.…”