TCP Congestion Research Articles

Yinker: A flexible BBR to achieve the high-throughput and low-latency data transmission over Wi-Fi and 5G networks

The popularization and development of Wi-Fi and 5G networks have introduced new applications requiring high data rates and low latency. However, the vast random packet loss caused by mobility and channel conditions in wireless networks can worsen the performance of traditional TCP congestion control algorithms. Therein, BBR [1] was proposed in 2016, and claims to operate at the optimum point. BBR strives to match the congestion window to the bandwidth-delay product, calculated from the measured bottleneck bandwidth and round trip times. Through simulations, we found that BBR suffers from different degrees of throughput degradation for different loss rates. The improved BBR, Yinker, is proposed to address this issue. Precisely, we dynamically adjust BBR’s pacing_gain based on the network conditions, including loss rate and congestion degree. We have evaluated Yinker in both real-world environments and trace-based emulations and compared its performance with different BBR variants and state-of-the-art schemes, including Cubic, Verus, and Copa. On average, TCP D*, BBR v2, Copa, and BBR have 4.24×, 2.34×, 2.01×, and 1.48× lower throughput compared to Yinker, respectively. This outstanding delay performance comes at little cost in latency. For instance, compared to TCP D* (which achieves the lowest latency), Yinker’s latency is only about 3% more.

Introducing a New TCP Variant for UAV networks following comparative simulations

The Transmission Control Protocol (TCP) is a reliable, connection oriented, congestion control mechanism, currently utilized the majority of both wired and wireless networks at the transport layer. An important function of TCP is the network congestion control mechanism; it governs the packet transmission rate and us enables the protocol to respond to congestion signals. Considering the wide spectrum of requirements stemming from unique network and channel characteristics, there exist numerous variants of TCP. While is commonly utilized in applications requiring reliable and ordered reception of packets, the standard TCP congestion control mechanism demonstrates poor performance in high-mobility wireless networking scenarios, as high mobility implies unreliable radio links and consecutive re-transmissions. In this paper we performed a comparative analysis of a spectrum of TCP variants, with the ultimate goal of deriving best practices to support real-time, yet reliable communication in high-mobility scenarios, with a special focus on aerial mobile ad hoc networks composed of Unmanned Aerial Vehicles (UAVs). Following the findings of this simulation evaluation, we introduce a new TCP variant for Flying Ad hoc Networks (FANETs), named Swarm HTCP (S-HTCP), which is shown to outperform the other variants in such network conditions.

LDM-satellite: A new scheme for packet loss classification over LEO satellite network

The packet loss classification has always been a hot and difficult issue in TCP congestion control research. Compared with the terrestrial network, the probability of packet loss in LEO satellite network increases dramatically. What's more, the problem of concept drifting is also more serious, which greatly affects the accuracy of the loss classification model. In this paper, we propose a new loss classification scheme based on concept drift detection and hybrid integration learning for LEO satellite networks, named LDM-Satellite, which consists of three modules: concept drift detection, lost packet cache and hybrid integration classification. As far, this is the first paper to consider the influence of concept drift on the loss classification model in satellite networks. We also innovatively use multiple base classifiers and a naive Bayes classifier as the final hybrid classifier. And a new weight algorithm for these classifiers is given. In ns-2 simulation, LDM-Satellite has a better AUC (0.9885) than the single-model machine learning classification algorithms. The accuracy of loss classification even exceeds 98%, higher than traditional TCP protocols. Moreover, compared with the existing protocols used for satellite networks, LDM-Satellite not only improves the throughput rate but also has good fairness.

A TCP Acceleration Algorithm for Aerospace-Ground Service Networks

The transmission of satellite payload data is critical for services provided by aerospace ground networks. To ensure the correctness of data transmission, the TCP data transmission protocol has been used typically. However, the standard TCP congestion control algorithm is incompatible with networks with a long time delay and a large bandwidth, resulting in low throughput and resource waste. This article compares recent studies on TCP-based acceleration algorithms and proposes an acceleration algorithm based on the learning of historical characteristics, such as end-to-end delay and its variation characteristics, the arrival interval of feedback packets (ACK) at the receiving end and its variation characteristics, the degree of data packet reversal and its variation characteristics, delay and jitter caused by the security equipment's deep data inspection, and random packet loss caused by various factors. The proposed algorithm is evaluated and compared with the TCP congestion control algorithms under both laboratory and ground network conditions. Experimental results indicate that the proposed acceleration algorithm is efficient and can significantly increase throughput. Therefore, it has a promising application prospect in high-speed data transmission in aerospace-ground service networks.

Aggregating Without Bloating: Hard Times for TCP on Wi-Fi

Since the definition of the bufferbloat phenomenon, several Linux kernel modules have been introduced in its TCP/IP stack, and there is a lack of experimental studies on their effects when coupled with WLAN technologies, in particular, IEEE 802.11n and IEEE 802.11ac. One essential algorithm introduced is named TCP Small Queues (TSQ) and has the role of limiting the number of packets that a TCP socket can enqueue in the stack, waiting for the physical layer to send the packets before enqueueing extra data. A second significant TCP algorithm is named TCP Pacing (TP) and regulates the pace used by the socket to enqueue packets in the stack, regulating the formation of bursts of data. These mechanisms affect the frame aggregation logic on WLAN networks and compromise the throughput-latency tradeoff of all the TCP variants. This paper presents an experimental evaluation of these techniques investigating the wireless network performance of several TCP congestion control variants under the presence of different TSQ and TP policies, modeling also their interaction.

Innovative Practice of Music Education in the Universities in the Context of 5G Network.

Music education is among the most significant subjects covered in providing high-quality education in Chinese universities and colleges. Music education is critical to providing high-quality education to students. It contributes significantly to the development of students' creative motivation, inventive capacity, and personality development. Music education provides excellent outcomes in music instruction and fosters students' original thinking and comprehensive abilities, and therefore supports the overall development of high-quality education. With the development of the educational system, it is becoming more vital to teach students a high level of musical literacy. With the advent of 5G mobile communication, it will become one of the core technologies in Chinese music education, providing an innovative framework for music education. In this study, a novel music education model is proposed for the development of music education using the GTZAN dataset which is comprised of 100 distinct specimens for every genre and ten various kinds of music. The dataset is normalized to prepare it for further processing and the characteristics of the song are retrieved using a technique called spectrum-based feature extraction (SBF). Bi-recurrent neural networks (Bi-RNN). are used to classify objects in space. An improved TCP congestion control algorithm (ITCCA) is proposed for efficient data transmission between the 5G networks. To optimize the performance of the transmission protocol, the honey bee optimization algorithm is employed. The performance of the proposed model is examined and contrasted with that of the currently used approaches. The proposed model shows high performance in terms of throughput, average delay, and packet delivery ratio. The model has the potential to successfully integrate 5G technologies and music education and provide the students with rich and diversified teaching materials and flexible instructional formats.

Choosing a Proper Control Strategy for Multipath Transmission in Industry 4.0 Applications

The industrial applications, especially in the context of communication-intensive Industry 4.0, require high throughput, good robustness, and low latency in the data exchange among the system components. These objectives can be achieved through multipath transfer. In order to overcome the restrictions of the primary networking protocols with respect to engaging multiple network interfaces, a new version of transmission control protocol (TCP) has recently been standardized—multipath TCP (MPTCP). Still though, only one type of congestion control algorithm is used to manage the flow of data despite various choices available in the modern operating systems, which leads to nonoptimal transfer dynamics. In this article, the influence of TCP congestion control algorithms, operating at the path and MPTCP level, on the transmission quality in industrial applications is examined. The conducted experiments involving physical devices and real, public networks lead to a set of recommendations for the algorithm selection so that the premises of robustness and efficiency are met. The default algorithm configuration is found ill-suited, yielding priority to nonhomogeneous settings. Only a few algorithms can maintain the benefits of multiplicity for an on-going transmission when the transfer conditions at one or more paths deteriorate. Motivated by the deficiency of common solutions, a new MPTCP congestion control algorithm has been developed. While retaining their good properties in the general networking context, this new–robust–algorithm yields improved performance in the uncertain, industrial communication environment.

MmS-TCP: Scalable TCP for Improving Throughput and Fairness in 5G mmWave Networks.

The millimeter-wave (mmWave) band, which can provide data rates of multi-gigabits per second, could play a major role in achieving the throughput goals of 5G networks. However, the high-bandwidth mmWave signal is susceptible to blockage by various obstacles, which results in very large and frequent degradation in the quality of the received signals. TCP, the most representative transport layer protocol, suffers from significant performance degradation due to the very dynamic channel conditions of the mmWave signal. Therefore, in this paper, we propose a congestion control algorithm that guarantees sufficient throughput in 5G mmWave networks and that does not significantly worsen TCP fairness. The proposed algorithm, which is a modification of Scalable TCP (S-TCP) that is designed for high-speed networks, provides a more stable performance than the existing TCP congestion control algorithm in mmWave networks through simple modifications. In various simulation experiments that considered the actual mobile user environment, the proposed mmWave Scalable TCP (mmS-TCP) algorithm demonstrated throughput up to 2.4 times higher than CUBIC TCP in single flow evaluation, and the inter-protocol fairness index when competing with CUBIC flow significantly improved from 0.819 of S-TCP to 0.9733. Moreover, the mmS-TCP algorithm reduced the number of duplicated ACKs by 1/4 compared with S-TCP, and it improved the average total throughput and intra-protocol fairness simultaneously.

Performance analysis of congestion control protocols over 5G mmWave links

Millimeter wave technology is an important basic technology in 5G applications. It has higher bandwidth and lower delay. But high dynamics of the channel poses a major challenge to the design of congestion control protocols. Based on ns3, this paper conducts extensive experiments on the TCP congestion control protocols, and analyzes the transmission performance of different protocols over 5G millimeter wave links in detail. It will lay a foundation for further research on congestion control protocols suitable for 5G millimeter wave communications.

Revisiting legacy high-speed TCP congestion control variants: An optimisation-theoretic analysis of multi-mode TCP

We revisit the problem of link capacity under-utilisation in TCP Congestion Control (TCP-CC) when working in High-Bandwidth-Delay-Product (High-BDP) networks. We approach this problem using a multi-mode approach and propose TCP-Gentle as an example of TCP-CC that uses this approach. While General Additive Increase Multiplicative Increase (GAIMD) congestion control algorithms received a lot of attention in the literature, little was mentioned about modelling multi-mode GAIMD. To this aim, we provide a tractable optimisation-theoretic model for TCP-Gentle which can be generalised to any multi-mode GAIMD. We show through analysis, simulation, and real-test-bed experiments of single flow, double flow, and single flow with background web traffic, that the proposed TCP-Gentle is competitive with existing TCP variants. Particularly, under certain assumption, TCP-Gentle can outperform TCP-YeAH in terms of fairness to TCP-NewReno. Besides, the proposed TCP-Gentle is more gentle to network; it maintains minimal average queues of less than 1.5% of pipe’s BDP, and reassembles to a great extent a highly-concave congestion window.

Throughput Analysis for TCP Over the FSO-Based Satellite-Assisted Internet of Vehicles

Free-space optics (FSO)-based, low-earth orbit (LEO) satellite-assisted Internet of vehicles (IoVs) has recently attracted research efforts worldwide. For the majority of Internet applications, transmission control protocol (TCP) is one of the most widely deployed protocols. Nevertheless, a complete understanding of the TCP performance over FSO-based satellite-assisted IoVs networks is still lacking in the literature. This is a highly challenging issue, as a consequence of the non-trivial interdependencies between transmission losses caused by the dynamics/uncertainty of atmospheric last-mile channels and the buffer overflows on the Internet, as well as their joint impact on the TCP congestion control mechanisms. This paper offers a comprehensive analytical framework that allows assessing the TCP throughput performance under the joint impact of congestion losses occurred in the Internet and transmission errors at the last-mile link from the satellite-to-vehicle. Internet access for emerging unmanned aerial vehicles (UAVs) is considered while the analysis can be also applied to other kinds of vehicles, which, in fact, require simpler channel models. The link-layer incremental redundancy hybrid automatic repeat request (IR-HARQ) protocol is employed at satellite-to-UAV channels to further improve the TCP throughput performance. Numerical results provide non-trivial findings and insights for the TCP performance in satellite/FSO-based IoVs, in which optimal UAV’s parameters are chosen to maximize the TCP throughput.

RAN Information-Assisted TCP Congestion Control Using Deep Reinforcement Learning With Reward Redistribution

In this paper, we aim to propose a novel transmission control protocol (TCP) congestion control method from a cross-layer-based perspective and present a deep reinforcement learning (DRL)-driven method called DRL-3R (DRL for congestion control with Radio access network information and Reward Redistribution) so as to learn the TCP congestion control policy in a superior manner. In particular, we incorporate the RAN information to timely grasp the dynamics of RAN, and empower DRL to learn from the delayed RAN information feedback potentially induced by several consecutive actions. Meanwhile, we relax the implicit assumption [that the feedback to one specific action returns at a round-trip-time (RTT) after the action is applied] in previous researches, by redistributing the rewards and evaluating the merits of actions more accurately. Experiment results show that besides maintaining a reasonable fairness, DRL-3R significantly outperforms classical congestion control methods (e.g., TCP Reno, Westwood, Cubic, BBR and DRL-CC) on network utility by achieving a higher throughput while reducing delay in various network environments.

Good Learning, Bad Performance: A Novel Attack Against RL-Based Congestion Control Systems

Reinforcement Learning (RL) has been applied to solve decision-making problems in computer network designs, especially in TCP congestion control. As RL-based congestion control methods enable powerful learning abilities, it achieves competitive performance and adaptiveness advantages over the traditional methods. However, RL-based systems suffer from adversarial attacks that generate perturbations to significantly degrade the performance. In this paper, we conduct a comprehensive study of adversarial attacks against RL-based congestion control systems. Unlike the state-of-the-art adversarial attacks on images where an attacker can easily obtain the input states to introduce perturbations, the attacker cannot directly obtain the input states in congestion control settings that are only available to the agents. It is challenging to add effective perturbations without knowing the input states for RL-based congestion control models. To solve the challenge, we develop an adversarial attack to estimate states of the target agent, craft adversarial perturbations, and apply the generated perturbations in an automated fashion. We evaluate how our adversarial attack affects the target agent’s decision-making process. Our experiments illustrate that our attack can effectively reduce about 50% average throughput while increasing more than 36x latency and 45% packet loss rate.

Enabling Robust DRL-Driven Networking Systems via Teacher-Student Learning

The past few years have witnessed a surge of interest towards deep reinforcement learning (DRL) in computer networks. With extraordinary ability of feature extraction, DRL has the potential to re-engineer the fundamental resource allocation problems in networking without relying on pre-programmed models or assumptions about dynamic environments. However, such black-box systems suffer from poor robustness, showing high performance variance and poor tail performance. In this work, we propose a unified Teacher-Student learning framework that harnesses rich domain knowledge to improve robustness. The domain-specific algorithms, less performant but more trustable than DRL, play the role of teachers providing advice at critical states; the student neural network is steered to maximize the expected reward as usual and mimic the teacher’s advice meanwhile. The Teacher-Student method comprises of three modules where the confidence check module locates wrong decisions and risky decisions, the reward shaping module designs a new updating function to stimulate the learning of student network, and the prioritized experience replay module to effectively utilize the advised actions. We further implement our Teacher-Student framework in existing video streaming (Pensieve), load balancing (DeepLB), and TCP congestion control (Aurora). Experimental results manifest that the proposed approach reduces the performance standard deviation of DeepLB by 37%; it improves the 90th, 95th, and 99th tail performance of Pensieve by 7.6%, 8.8%, and 10.7% respectively; and it accelerates the growth rate of Aurora by 2x at the initial stage, and achieves a more stable performance in dynamic environments.

Improving Throughput of Transmission Control Protocol Using Cross Layer Approach

Most of the internet users connect through wireless networks. Major part of internet traffic is carried by Transmission Control Protocol (TCP). It has some design constraints while operated across wireless networks. TCP is the traditional predominant protocol designed for wired networks. To control congestion in the network, TCP used acknowledgment to delivery of packets by the end host. In wired network, packet loss signals congestion in the network. But rather in wireless networks, loss is mainly because of the wireless characteristics such as fading, signal strength etc. When a packet travels across wired and wireless networks, TCP congestion control theory faces problem during handshake between them. This paper focuses on finding this misinterpretation of the losses using cross layer approach. This paper focuses on increasing bandwidth usage by improving TCP throughput in wireless environments using cross layer approach and hence named the proposed system as CRLTCP. TCP misinterprets wireless loss as congestion loss and unnecessarily reduces congestion window size. Using the signal strength and frame error rate, the type of loss is identified and accordingly the response of TCP is modified. The results show that there is a significant improvement in the throughput of proposed TCP upon which bandwidth usage is increased.

Comparison of TCP SIAD and TCP BBR Congestion Control in Simulated 5G Networks

Abstract 5G cellular networks have introduced a completely novel air interface called New Radio (NR). This technology delivers numerous benefits compared to previous generations, including significantly higher peak data rates. However, due to the propagation properties of the frequencies used in NR, the volatility of the available downlink capacity also increases. In this paper, we study two TCP congestion control algorithms which are designed to be able to quickly utilize sudden increases in available capacity. We present an implementation of TCP SIAD in the ns-3 open source network simulator and compare its performance with TCP BBR using the mmWave module of the simulator.

DeepCC: Multi-Agent Deep Reinforcement Learning Congestion Control for Multi-Path TCP Based on Self-Attention

With the development of the Internet of Things (IoT) and 5G, there are ubiquitous smart devices and network functions providing emerging network services efficiently and optimally through building many network connections based on WiFi, LTE/5G, Ethernet, and etc. The Multipath TCP (MPTCP) protocol that enables these devices to establish multiple paths for simultaneous data transmission, has been a widely used extension of standard TCP in smart devices and network functions. On the other hand, more heavy and time-varying traffic loads are generated in an MPTCP network, so that an efficient congestion control mechanism that schedules the traffic between multiple subflows and avoids congestion is highly required. In this paper, we propose a decentralized learning approach, DeepCC, to adapt to the volatile environments and realize the efficient congestion control. The Multi-Agent Deep Reinforcement Learning (MADRL) is used to learn a policy of congestion control for each subflow according to the real-time network states. To deal with the problem of the fixed state space and slow convergence, we adopt two self-attention mechanisms to receive the states and train the policy, respectively. Due to the asynchronous design of DeepCC, the learning process will not introduce extra delay and overhead on the decision-making process. Experiment results show that DeepCC consistently outperforms the well-known heuristic method and DRL-based MPTCP congestion control method in terms of goodput and jitter. Besides, DeepCC with the attention mechanism reduces convergence time by about 50% and increase goodput by about 80% compared with the commonly used structures of neural networks.

An Intelligent TCP Congestion Control Method Based on Deep Q Network

To optimize the data migration performance between different supercomputing centers in China, we present TCP-DQN, which is an intelligent TCP congestion control method based on DQN (Deep Q network). The TCP congestion control process is abstracted as a partially observed Markov decision process. In this process, an agent is constructed to interact with the network environment. The agent adjusts the size of the congestion window by observing the characteristics of the network state. The network environment feeds back the reward to the agent, and the agent tries to maximize the expected reward in an episode. We designed a weighted reward function to balance the throughput and delay. Compared with traditional Q-learning, DQN uses double-layer neural networks and experience replay to reduce the oscillation problem that may occur in gradient descent. We implemented the TCP-DQN method and compared it with mainstream congestion control algorithms such as cubic, Highspeed and NewReno. The results show that the throughput of TCP-DQN can reach more than 2 times of the comparison method while the latency is close to the three compared methods.

Model-Agnostic and Efficient Exploration of Numerical Congestion Control State Space of Real-World TCP Implementations

The significant impact of TCP congestion control on the Internet highlights the importance of testing congestion control algorithm implementations (CCAIs) in various network environments. Many CCAI testing problems can be solved by exploring the numerical state space of CCAIs, which is defined by a group of numerical (and nonnumerical) state variables of the CCAIs. However, the current practices for automated numerical state space exploration are either limited by the approximate abstract CCAI models or inefficient due to the large space of network environment parameters and the complicated relation between the CCAI states and network environment parameters. In this paper, we propose an automated numerical state space exploration method, called ACT, which leverages the model-agnostic feature of random testing and greatly improves its efficiency by guiding random testing under the feedback iteratively obtained in a test. Our experiments on five representative Linux TCP CCAIs show that ACT can more efficiently explore a large numerical state space than manual testing, undirected random testing, and symbolic execution based testing, while without requiring an abstract CCAI model. ACT detects multiple design and implementation bugs of these Linux TCP CCAIs, including some new bugs not reported before.

On the Detection of Low-Rate Denial of Service Attacks at Transport and Application Layers

Distributed denial of service (DDoS) attacks aim to deplete the network bandwidth and computing resources of targeted victims. Low-rate DDoS attacks exploit protocol features such as the transmission control protocol (TCP) three-way handshake mechanism for connection establishment and the TCP congestion-control induced backoffs to attack at a much lower rate and still effectively bring down the targeted network and computer systems. Most of the statistical and machine/deep learning-based detection methods proposed in the literature require keeping track of packets by flows and have high processing overheads for feature extraction. This paper presents a novel two-stage model that uses Long Short-Term Memory (LSTM) and Random Forest (RF) to detect the presence of attack flows in a group of flows. This model has a very low data processing overhead; it uses only two features and does not require keeping track of packets by flows, making it suitable for continuous monitoring of network traffic and on-the-fly detection. The paper also presents an LSTM Autoencoder to detect individual attack flows with high detection accuracy using only two features. Additionally, the paper presents an analysis of a support vector machine (SVM) model that detects attack flows in slices of network traffic collected for short durations. The low-rate attack dataset used in this study is made available to the research community through GitHub.