A Novel Congestion Control Algorithm Based on Adaptive Traffic Delay

Despite many years of improvements to it, TCP still suffers from an unsatisfactory performance. For services dominated by short flows (e.g., web search and e-commerce), TCP suffers from the flow startup problem and cannot fully utilize the available bandwidth in the modern Internet: TCP starts from a conservative and static initial window (IW, 2-4 or 10), while most of the web flows are too short to converge to the best sending rate before the session ends. For services dominated by long flows (e.g., video streaming and file downloading), the congestion control (CC) scheme manually and statically configured might not offer the best performance for the latest network conditions. To address these two challenges, we propose TCP-RL, which uses reinforcement learning (RL) techniques to dynamically configure IW and CC in order to improve the performance of TCP flow transmission. Basing on the latest network conditions observed at the server side of a web service, TCP-RL dynamically configures a suitable IW for short flows through group-based RL, and dynamically configures a suitable CC scheme for long flows through deep RL. Our extensive experiments show that for short flows, TCP-RL can reduce the average transmission time by about 23%; and for long flows, compared with the performance of 14 CC schemes, TCP-RL's performance ranks top 5 for about 85% of the 288 given static network conditions, whereas for about 90% of conditions, its performance drops by less than 12% compared with that of the best-performing CC schemes for the same network conditions.

RaceCC: A rapidly converging explicit congestion control for datacenter networks

Congestion control has been emphasis of network research. The researchers proposed some algorithm of congestion control in order to continuously reduce the packet dropping possibility, and improve the link utilization ratio. First the paper directed the user by comparing some typical AQM control algorithms, enabled them optimize the network transport performance using more reasonable control algorithm in different network environment. Finally the paper proposed a new congestion control algorithm,which by predicting the instantaneous queue length of next time to decide congestion. It can find the congestion at the earliest time and make the best of network resource. The simulation result shows the proposal can detect the congestion in time in the changing network condition, and make the output smooth.

Deeper investigation is needed for several open issues related to congestion and admission control within the field of the high speed B‐ISDN network with ATM protocol. The system must be more effective in resource management and robust enough to cope with changes in traffic environment, which nowadays is characterised by sophisticated multimedia services. In this paper a new admission and congestion control strategy is proposed. It originates from the Fast Buffer Reservation (FBR) scheme, a virtual circuit acceptance algorithm proposed by Prof. Turner, and enhances its effectiveness, thus resulting in a complete approach for bandwidth management. The scheme is called Modified FBR (M‐FBR). The use of the FBR method, in fact, when based on the powerful Excess Probability Criterion (EPC), seemed to be impeded by the occurrence of lock out problems. By adopting a new kind of service at the buffer exit, the new mechanism offers the required flexibility and robustness, as well as ease of control, while preserving burst integrity and exhibiting a high burst‐served to burst‐sent ratio.

Machine-to-machine (M2M) communications are bringing new challenges to congestion control in the Internet of Things. One key issue is to facilitate the proper functioning of a wide range of M2M applications with drastically different throughput demands. Traditional Internet congestion control algorithms aim at sharing bandwidth among traffic flows equally, limiting their suitability for M2M communications. To maintain comparable levels of QoS of heterogeneous M2M applications when congestion is present, we propose a distributed congestion control algorithm which allocates transmission rates to M2M flows in proportion to their demands, through the use of a simple technique which we call “proportional additive increase”. To ease M2M application development, we make a further attempt to stabilize the throughputs of M2M flows controlled by the algorithm. We present simulation results to illustrate the effectiveness of the algorithm in achieving the desired rate allocation, as well as the challenge we face in stabilizing throughputs.

In the modern cellular data networks, with the tremendous amount of mobile devices, the constantly decreasing prices of the hardware and according the trends worldwide, provisioning of the data service and its quality will soon become an impossible task. Since current mobile cellular standards are unable to serve and meet the needs of this increase, the network will eventually become congested. As we are investigating data networks, they are based on the TCP protocol. TCP has various congestion control schemes used in both wired and wireless network, such as Cubic, Vegas, Reno, etc. Most modern cellular systems have already been migrated to 3G and 4G, so it could be very useful to identify which algorithm performs best in LTE (4G) wireless network. As most of the platforms for mobile devices, such as phones and tablets are unix/linux based and the current default algorithm in Linux is Cubic, the focus in this article will be on real-time modification and testing of the congestion control parameters of TCP CUBIC.

Aiming at the network congestion control problem, a congestion control algorithm based on Actor-Critic reinforcement learning model is designed. Through the genetic algorithm in the congestion control strategy, the network congestion problems can be better found and prevented. According to Actor-Critic reinforcement learning, the simulation experiment of network congestion control algorithm is designed. The simulation experiments verify that the AQM controller can predict the dynamic characteristics of the network system. Moreover, the learning strategy is adopted to optimize the network performance, and the dropping probability of packets is adaptively adjusted so as to improve the network performance and avoid congestion. Based on the above finding, it is concluded that the network congestion control algorithm based on Actor-Critic reinforcement learning model can effectively avoid the occurrence of TCP network congestion.

A priority scheduling algorithm with double threshold and a traffic control mechanism with hysteresis effect are proposed. The double threshold priority scheduling is studied based on the minimum laxity threshold (MLT) and the queue length threshold (QLT). The hysteresis effect traffic control is specified by the hysteresis effect QLT and traffic rate control. According to the simulation results, it can be shown that the proposed dynamic priority scheduling has a better processing performance than the existing MLT and QLT algorithm. The results also demonstrate that the QLT with hysteresis effect and traffic rate control with the hysteresis effect enhance the performance in comparison with those of single threshold.

In recent years, hybrid congestion control (CC) algorithms that combine rule-based CC and learning-based CC have gained significant attention. They incorporate the fast adaption ability of learning-based CC and the stability of rule-based CC, tending to select the better-performing rate based on the network feedback. However, the practical implementation of such algorithms has revealed primary issues. Specifically, they require both CCs to run alternately, which results in a poorly performing CC continuing to run in the network. Moreover, hybrid CCs cannot converge to the optimal rate when both CCs perform poorly. This paper proposes Anole to address these issues. Anole has three main algorithmic contributions: 1) Anole always selects the better-performing CC, 2) Anole temporarily deprecates the consistently underperforming CC, 3) when both CCs pervform poorly, Anole infers the optimal sending rate based on the network feedback. We carry out comprehensive experiments in both emulated and real-world wired networks, as well as in real-world WiFi networks, to assess the performance of Anole. The experiment results demonstrate that Anole achieves approximately 6% higher throughput in real-world links and 34% lower delay in the 48Mbps link compared to the state-of-the-art CC. Anole also exhibits superior performance in adaptability and fair convergence.

Traditional congestion control algorithms rely on various model-based methods to improve the end-to-end (E2E) performance of packet transmission. The resulting decisions quickly become less effective amid the dynamics of network conditions. In order to perform congestion control adaptively, reinforcement learning (RL) can be adopted to continuously learn the optimal strategy from the network environment. Oftentimes, the reward of such a learning problem is a weighted sum of multiple E2E performance metrics, such as throughput, delay, and fairness. Unfortunately, those weights can be only manually tuned based on extensive experiments. To address this issue, in this paper, we design a constrained RL algorithm for congestion control named CRL-CC to adaptively tune those weights, with the objective of effectively improving the overall E2E packet transmission performance. In particular, the multi-objective optimization problem is firstly formulated as a constrained optimization problem. Then, the Lagrangian relaxation method is leveraged to transform the constrained optimization problem into a single-objective optimization problem, which is solved by designing a multi-objective reward function with Lagrangian multipliers. Extensive experiments based on OpenAI-Gym show that the proposed CRL-CC algorithm can achieve higher overall performance in various network conditions. In particular, the CRL-CC algorithm outperforms the benchmark algorithm on Pantheon by 21.7%, 27.4%, and 5.3% in throughput, delay, and fairness, respectively.

Chapter 5 - Congestion Control in High-Speed Networks

A key component in the QoS architecture of the network is congestion control algorithms. Active congestion control is a new congestion control technology based on active network. it has small feedback delay, control nimble and high utility of network resource. A new active congestion control algorithm based on secondary routing was put forward, By the study of congestion detection, secondary routing search and choosing, multi-path distributary and control, a new congestion control method by packets redirection on active node was explored. The performance of this algorithm were validated by NS2, the following works of this algorithm was given out.

In 2016, Google proposed the bottleneck bandwidth round-trip propagation time (BBR) as a new congestion control algorithm for improving network speed. As BBR has different behavioral characteristics from loss-based congestion control, it must ensure the coexistence between congestion control algorithms. However, throughput imbalance has been reported in a competition between BBR and loss-based congestion control. In this paper, we propose an opponent congestion control identification model of BBR (OI-BBR) with machine learning to improving inter-protocol fairness. In the emulation experiments, we confirmed that OI-BBR based on the proposed model improved the inter-protocol fairness by operating differently according to the opponent congestion control algorithms.

Congestion control is a fundamental mechanism for TCP protocol, which has been extensively studied in the past three decades. However, our experimental evaluations show that the state-of-art congestion control algorithms such as Cubic and BBR are far from optimal: they have unresolved issues such as insufficient usage of available bandwidth, inadaptable to dynamic bandwidth variants, and compromising on one or more performance dimensions. To address these challenges, we propose a novel congestion control mechanism called NeuRoc that coordinately uses online changepoint detection and deep reinforcement learning (DRL) technique to generate the optimal congestion control policy, which allows TCP operating at Kleinrock ’s optimal operation point to achieve fully bandwidth usage and low latency. To address the practical issues of deploying the deep learning based congestion control mechanism, we propose a cold-started training and deployment framework to reduce the cost of bootstrap. We implement NeuRoc on an emulation platform which connects to the Linux network protocol stack through virtual network interfaces. Extensive experiments show that NeuRoc achieves the best throughput-latency tradeoff compared with the state-of-the-arts in a variety of scenarios.

The traditional congestion control mechanism, i.e., TCP, when applied in mobile ad hoc networks (MANET), encounters a number of new challenges, such as wireless link error, medium contention and frequent route failures. Very poor performance of TCP in MANET has been reported in many recent studies. We focus on the problems resulting from medium contention and propose a novel rate based end-to-end congestion control scheme (RBCC). We first illustrate that, under the impact of medium contention, a window based congestion control algorithm is unstable and hence may not be appropriate for MANET because the optimum congestion window size is very small and may even be less than one, i.e., the source should send less than one packet in one round trip time (RTT). Based on the novel use of channel busyness ratio, which, we show, is an accurate sign of the network utilization and congestion status, a new rate control scheme has been proposed to support the transport service in MANET efficiently and reliably. In RBCC, a sublayer consisting of a leaky bucket is added under TCP to control the sending rate based on the network layer feedback at the bottleneck node. Extensive simulations show that our scheme significantly outperforms traditional TCP in terms of channel utilization, delay and fairness.