TCP Congestion Research Articles

TCP-NeuRoc: Neural Adaptive TCP Congestion Control With Online Changepoint Detection

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.

A Comprehensive Overview of TCP Congestion Control in 5G Networks: Research Challenges and Future Perspectives.

In today’s data networks, the main protocol used to ensure reliable communications is the transmission control protocol (TCP). The TCP performance is largely determined by the used congestion control (CC) algorithm. TCP CC algorithms have evolved over the past three decades and a large number of CC algorithm variations have been developed to accommodate various network environments. The fifth-generation (5G) mobile network presents a new challenge for the implementation of the TCP CC mechanism, since networks will operate in environments with huge user device density and vast traffic flows. In contrast to the pre-5G networks that operate in the sub-6 GHz bands, the implementation of TCP CC algorithms in 5G mmWave communications will be further compromised with high variations in channel quality and susceptibility to blockages due to high penetration losses and atmospheric absorptions. These challenges will be particularly present in environments such as sensor networks and Internet of Things (IoT) applications. To alleviate these challenges, this paper provides an overview of the most popular single-flow and multy-flow TCP CC algorithms used in pre-5G networks. The related work on the previous examinations of TCP CC algorithm performance in 5G networks is further presented. A possible implementation of TCP CC algorithms is thoroughly analysed with respect to the specificities of 5G networks, such as the usage of high frequencies in the mmWave spectrum, the frequent horizontal and vertical handovers, the implementation of the 5G core network, the usage of beamforming and data buffering, the exploitation of edge computing, and the constantly transmitted always-on signals. Moreover, the capabilities of machine learning technique implementations for the improvement of TCPs CC performance have been presented last, with a discussion on future research opportunities that can contribute to the improvement of TCP CC implementation in 5G networks. This survey paper can serve as the basis for the development of novel solutions that will ensure the reliable implementation of TCP CC in different usage scenarios of 5G networks.

Understanding Degradation Attack and TCP Performance in Next Generation Passive Optical Network

The next-generation gigabit passive optical network (NG-PON) is a promising solution for network growth and expansion. The NG-PON aims to connect more users as fiber’s deployment to the home (FTTH) initiative takes place. The network needs urgent attention to its security establishment, which is currently exposed to a degradation attack. During the attack, already available TCP congestion control is invoked to handle collision. The interconnection of TCP in the transport layer and attack in the MAC layer is investigated. During the attack, a reduction of 63% of total throughput is detected. A variation of the number of ONU also determines the competing nature of TCP, which inherently reduces the maximum throughput of 4%.

Performance evaluation of massively parallel and high speed connectionless vs. connection oriented communication sessions

In this paper we focus on the fast communication issues of the Big Data processing tasks shared between High Performance Computing systems. In our performance evaluation framework we designed and developed two traffic measurement tools in order to answer some theoretical questions related to congestion control in practice. The first one is based on iperf and tcpdump softwares to capture data flows of TCP and UDP sessions. Classification aspects of the measurement cases were: homogeneity of the traffics, number of parallel communication sessions, and implementation types of the TCP congestion control algorithm. Dozens of parallel traffic scenarios were executed in a dumbbell topology to evaluate effects of the massively parallel communication sessions in wireline local and metropolitan area networks. Since we found that connection oriented data transfer sessions have limited performance features during communication, we implemented a second communication tool named Fast Manager of File Transfer (FMFT). This application with transfer rate monitoring and regulation capability is based on parallel connectionless data transfer sessions supervised by a common connection oriented control session and provides better transfer rate than the classical file transfer mechanisms using TCP services. Methodology of the statistical analysis and highlights of this heterogeneous parallel communication mechanism are explained, too.

Performance analysis of high-speed TCP protocols in LTE X2 handover under realistic operational conditions

Broadband mobile networks have rapidly evolved over the last years. The unique environment they operate, however, sets new challenges for the Transmission Control Protocol (TCP). In this paper, we study the impact of Long Term Evolution (LTE) handover on TCP performance with different TCP congestion control algorithms and in a wide range of networks settings, including different handover types, random channel errors and support of error correction at lower layers. Our results suggest that the congestion control algorithms employed by TCP variants during the LTE X2 handover have different impacts on TCP performance. In particular CUBIC, Scalable TCP and BIC-TCP achieve high goodput and fast window convergence in available channel capacity after LTE X2 handover. In environments involving packet drops that are not corrected at lower layers and single errors with high probability, all TCP variants exhibit very low performance levels. Finally, in terms of fairness, our results identify fairness issues among TCP flows with different Bandwidth Delay Products (BDP), with CUBIC offering the best results.

Intelligent Forwarding Strategy for Congestion Control Using Q-Learning and LSTM in Named Data Networking

Named data networking (NDN) is a future network architecture that replaces IP-oriented communication with content-oriented communication and has new features such as cache, multiple paths, and multiple sources. Services such as video streaming, to which NDN can be applied in the future, can cause congestion if data is concentrated on one of the nodes during high demand. To solve this problem, sending rate control methods such as TCP congestion control have been proposed, but they do not adequately reflect the characteristics of NDN. Therefore, we use reinforcement learning and deep learning to propose a congestion control method that takes advantage of multipath features. The intelligent forwarding strategy for congestion control using Q-learning and long short-term memory in NDN proposed in this paper is divided into two phases. The first phase uses an LSTM model to train a pending interest table (PIT) entry rate that can be used as an indicator to detect congestion by knowing the amount of data returned. In the second phase, it is forwarded to an alternative path that is not congestive via Q-learning based on the PIT entry rate predicted by the trained LSTM model. The simulation results show that the proposed method increases the data reception rate by 6.5% and 19.5% and decreases the packet drop rate by 7.3% and 17.2% compared to an adaptive SRTT-based forwarding strategy (ASF) and BestRoute.

Fair hop-by-hop interest rate control to mitigate congestion in named data networks

Data retrieval in Named Data Networking (NDN) is based on content names irrespective of their hosting location. The NDN architecture introduces original naming, forwarding and caching techniques to improve data delivery efficiency. These distinctive techniques make the TCP congestion control scheme not suitable for the NDN architecture. In particular, the in-network caching and the multi-path forwarding used in the NDN architecture lead to fluctuations in the Round Trip Time (RTT) measurements. This makes the Retransmission TimeOut (RTO) an unreliable congestion indicator. Thus, the congestion problem in Named Data Networks deserves to be reconsidered. In this paper, we present a new congestion control algorithm, called IRNA (Interest Rate Notification and Adjustment), to mitigate the congestion problem in Named Data Networks while maintaining per flow and per consumer fairness property. More specifically, IRNA is a hop-by-hop scheme where each router monitors the occupation level of its outgoing queues to detect in advance potential congestion events. When the queue size is above or below a preset threshold, the router sends explicit notifications to its downstream nodes to specify the appropriate interest rate. Hence, by cooperatively adjusting the Interest transmission rate of each node and for each flow, IRNA succeeded to provide a fair transmission. We evaluated the performance of our solution using ndnSIM simulations. We compared IRNA to state-of-the-art hop-by-hop congestion control solution and we proved that IRNA managed to provide a faster congestion reaction and to achieve per flow and per consumer throughput fairness while maintaining a good bandwidth usage.

Improved RTT Fairness of BBR Congestion Control Algorithm Based on Adaptive Congestion Window

To alleviate the lower performance of Transmission Control Protocol (TCP) congestion control over complex network, especially the high latency and packet loss scenario, Google proposed the Bottleneck Bandwidth and Round-trip propagation time (BBR) congestion control algorithm. In contrast with other TCP congestion control algorithms, BBR adjusted transfer data by maximizing delivery rate and minimizing delay. However, some evaluation experiments have shown that the persistent queues formation and retransmissions in the bottleneck can lead to serious fairness issues between BBR flows with different round-trip times (RTTs). They pointed out that small RTT differences cause unfairness in the throughput of BBR flows and flows with longer RTT can obtain higher bandwidth when competing with the shorter RTT flows. In order to solve this fairness problem, an adaptive congestion window of BBR is proposed, which adjusts the congestion window gain of each BBR flow in network load. The proposed algorithms alleviate the RTT fairness issue by controlling the upper limit of congestion window according to the delivery rate and queue status. In the Network Simulator 3 (NS3) simulation experiment, it shows that the adaptive congestion window of BBR (BBR-ACW) congestion control algorithm improves the fairness by more than 50% and reduces the queuing delay by 54%, compared with that of the original BBR in different buffer sizes.

A review on Queue Management Algorithms in Large Networks

The efficiency of Transmission Control Protocol (TCP) is significantly influenced by Congestion Control. Many scientists have widely researched and suggested several improvements to conventional TCP Congestion Control over the past centuries. This subject, however, still inspires scholarly and industrial research groups due to changes in demands for internet applications and the evolution of internet technologies. In this review paper, illustrate the most important features of the TCP and its mechanisms. The challenges and its effect on the performance of the network communication have also been explained. Several research groups have been working on the improvement of the quality of services of the network. This paper reviews some recent studies and argues their latest findings to have a better understanding. Where the important features of sub-sections of the TCP were presented. Then we showed the two main classes of the existing router queue management system. Performance metrics are used to show the behaviour of a network during data transmission. The results in this study showed the performance of ARED was be better than that of RED.

Mitigating TCP Protocol Misuse With Programmable Data Planes

This article proposes a new approach for detecting and mitigating the impact of misbehaving TCP end-hosts, specifically the Optimistic ACK attack, and Explicit Congestion Notification (ECN) abuse. In contrast to the state-of-the-art, we show that it is possible to mitigate such misbehavior leveraging emerging programmable data planes while not requiring any end-host or protocol modifications. A key challenge in doing so is to implement expressive, complex and stateful functions in the data plane within its restricted programming model. In this regard, we propose a security monitoring function that uses Extended Finite State Machine (EFSM) abstraction for monitoring stateful protocols in the data plane. We also design a mechanism for mapping a protocol’s EFSM to programmable data plane primitives. Our evaluation results demonstrate that our approach can fully or partially restore the throughput loss caused by misbehaving end-hosts that manipulate TCP congestion control through misinformation.

High-Potency Models of LDoS Attack Against CUBIC + RED

A TCP-targeted Low-rate denial of service (LDoS) attack exploits the vulnerabilities of TCP congestion control mechanism. CUBIC is the most widely used TCP congestion control algorithm today. CUBIC TCP increases the resilience against LDoS over traditional TCP. This paper explores high-potency patterns of the LDoS attack against CUBIC TCP under RED queue management scenario, and accordingly develops two attack models referring to maximizing attack potency. Theoretical analyses and extensive experiments are conducted to validate the proper function of the two models and evaluate their performance. Test results show that the two attack models can effectively throttle CUBIC TCP throughput. Under standard-configured network parameters, the number of TCP units damaged by one attack unit are up to about 21 and 26 respectively for our proposed two models, which represents an increase in attack potency of about 20%. And, our proposed models outperform the traditional attack model in terms of attack potency by at least 250%. In addition, with variations in different network parameters, these two models are still efficient and alternatively maximize the attack potency. Finally, an outline for the attack countermeasure is discussed. The present study offers a basis to explore new attack manners which may be exploited by attackers and excites defenders to develop new measurements against such attack.

A Trusted NUMFabric Algorithm for Congestion Price Calculation at the Internet-of-Things Datacenter

The important issues of network TCP congestion control are how to compute the link price according to the link status and regulate the data sending rate based on link congestion pricing feedback information. However, it is difficult to predict the congestion state of the link-end accurately at the source. In this paper, we presented an improved NUMFabric algorithm for calculating the overall congestion price. In the proposed scheme, the whole network structure had been obtained by the central control server in the Software Defined Network, and a kind of dual-hierarchy algorithm for calculating overall network congestion price had been demonstrated. In this scheme, the first hierarchy algorithm was set up in a central control server like Opendaylight and the guiding parameter B is obtained based on the intelligent data of global link state information. Based on the historical data, the congestion state of the network and the guiding parameter B is accurately predicted by the machine learning algorithm. The second hierarchy algorithm was installed in the Openflow link and the link price was calculated based on guiding parameter B given by the first algorithm. We evaluate this evolved NUMFabric algorithm in NS3, which demonstrated that the proposed NUMFabric algorithm could efficiently increase the link bandwidth utilization of cloud computing IoT datacenters.

Protocols for Transferring Bulk Data Over Internet: Current Solutions and Future Challenges

Big Data has gained interests in effectively capturing, storing, analysis and visualisation from wide range of scientific, economic and business communities and is frequently communicated over internet for various purposes among government and enterprise sectors sited at different locations. Several experiments and analyses have shown that currently employed applications and transport protocols in internet are not suitable for transferring such voluminous data because of not addressing requirements of low-access latency. This paper presents issues associated with the basic mechanism of legacy protocols in the context of high speed networks for transferring Big Data e.g. conservative TCP congestion control mechanism may result in minute utilisation of high bandwidth provisioning networks. We present state-of-art alternatives proposed in the literature to solve these problems in high speed networks. We compare several underlying emerging alternatives of TCP, UDP and multi-TCP-streams protocols over a number of comparison criteria e.g. protocol convergence, responsiveness etc., to handle communication of huge data. We note that these protocol alternatives have significant importance over fulfilling requirements of emerging data-intensive applications in high-speed networks. In addition, we discuss open research issues and challenges that can be explored as a source of motivation towards development and deployment of data-intensive applications in emerging networking technologies.

ABCSS: A novel approach for increasing the TCP congestion window in a network

Congestion control is a useful task to be taken into account in order to maintain a good traffic in a network. In the case of a TCP/IP (Transmission Control Protocol/Internet Protocol) network, congestion creates queues in routers and even packet rejection. It is the reason why research have been conducted to improve congestion control mechanisms. Nowadays, several TCP congestion algorithms have been implemented. However, with the technology evolution and the need in terms of traffic capacity, various mechanisms implemented by these algorithms have shown their limitations and still need to be improved. The aim of this paper is to develop a novel TCP congestion control approach based on two famous algorithms: Appropriate Byte Counting (ABC) and Slow Start (SS). The results obtained show that ABC increases the congestion window appropriately compared to Slow Start. However, the congestion window remains invariant during the first round trips, which causes the TCP burst phenomenon. The TCP splitting can cause buffer overflows and long queues. For theses shortcomings, we propose ABCSS as a novel approach that appropriately increase the TCP congestion window and minimize the TCP splitting problem. The simulations of ABCSS algorithm provide interesting results compared to some known algorithms.

An Improved Algorithm to Enhance the Performance of FAST TCP Congestion Control for Personalized Healthcare Systems

The development of personalized medical systems should be supported by a fast and stable network system. The FAST TCP network system is the appropriate support system for this purpose. However, when the FAST TCP is deployed, the static mapping selection method for protocol parameters is unable to guarantee the small queuing delay and fast convergence of the network simultaneously. By conducting theoretical analysis and simulation experiments, the relationships among FAST TCP protocol slow start condition, control law gain parameters, and FAST TCP system convergence rate were examined. To ensure the stability of the FAST TCP system and to select the smallest protocol parameters, an improved method to effectively accelerate the convergence velocity of the FAST TCP system is proposed in this study. In this method, the number of packets for staying in the buffer for FAST TCP connections was taken as the criterion of the slow start, and the gain parameter of the control law was dynamically adjusted according to the local information of each FAST TCP connection. Using this improved method, the FAST TCP system can achieve a stable and small queuing delay, whilst the FAST TCP system could converge quickly to the equilibrium point simultaneously.

, , , and

, , , and

STCC: A SDN‐oriented TCP congestion control mechanism for datacenter network

Due to characteristics of high bandwidth and low latency, datacenter networks ensure tremendous data could be transmitted in an efficient way. However, in many-to-one transmission scenarios, high concurrency of TCP flows aggravates network congestion and causes overflows in switches, seriously impairing network performance. To solve the problem, a TCP congestion control mechanism based on software-defined networking (STCC) is proposed. Without any modification in TCP stack, STCC monitors network performance through centralized control and global network view of SDN, employs a routing algorithm based on the minimum path bandwidth utilization rate to forward packets and uses different methods to adjust congestion windows of senders so that network congestion can be greatly mitigated. An experiment platform is built to carry out simulation tests for evaluating STCC, and the results show that under the same network conditions, STCC effectively reduces the number of retransmission timeout of senders and noticeably raises network throughput, compared with other congestion control algorithms.

FAMD: A Flow-Aware Marking and Delay-based TCP algorithm for datacenter networks

Datacenter networks have brought new challenges to meet datacenters’ unique performance requirements. Moreover, the many-to-one traffic pattern, which is quite common in datacenter networks, causes heavy network congestion and leads to TCP incast problem. To solve this problem, many TCP congestion control protocols have been proposed to mitigate the TCP incast. However, these protocols have more or less problems, including high computational complexity and implementation complexity. In this paper, we propose a flow-aware marking and delay-based TCP algorithm for datacenters, referred to as FAMD. FAMD is built on the delay-based protocols and leverages a DCTCP-like marking scheme to indicate the network congestion level and regulate the congestion window. Based on above mechanism, FAMD can maintain a low and stable transmission end-to-end delay. Furthermore, because of the disadvantage of fixed marking threshold, this paper shows how to choose an appropriate threshold dynamically and presents a flow-aware threshold adjustment algorithm, referred to FATA. Therefore, our FAMD can achieve a good generality for different scenarios. Our extensive evaluation shows that FAMD can gain a good throughput and low end-to-end delay.

TCP CERL+: revisiting TCP congestion control in wireless networks with random loss

In this paper, we analyze the performance of wireless networks subject to random loss. In this regard, we revisit our TCP Congestion Control Enhancement for Random Loss (CERL) mechanism that has been introduced earlier in a previous study. We call our new developed TCP variant as CERL Plus (CERL+). TCP CERL+ is a modification of TCP Reno at sender-side. TCP CERL+ is a new generation that works similarly as TCP CERL but its main idea is to use a dynamic threshold in terms of RTT. In doing this, we employ the average RTT and its minimum measurements made over the connection to estimate the queue length of the bottleneck link. As a result, we can use this queue length to evaluate the congestion status and distinguish it from the random loss status. This in turn would not reduce the window size leading to enhance the performance of the proposed algorithm in the sense of increasing the amount of data transmission. Additionally, CERL+ alleviates the congestion in the bottleneck obviously. In this paper, we present simulation experiment for TCP CERL+ considering a two-way transmission assuming a heavy load and wide range of random loss rates compared to TCP NewReno, TCP New Jersey + , TCP mVeno, TCP Westwood + , TCP Cubic and TCP YeAh using the network simulation ns-2. Simulation results prove that CERL+ outperforms other TCP variants and achieves excellent throughput gain.

Prophet: Toward Fast, Error-Tolerant Model-Based Throughput Prediction for Reactive Flows in DC Networks

As modern network applications ( e.g. , large data analytics) become more distributed and can conduct application-layer traffic adaptation, they demand better network visibility to better orchestrate their data flows. As a result, the ability to predict the available bandwidth for a set of flows has become a fundamental requirement of today’s networking systems. While there are previous studies addressing the case of non-reactive flows, the prediction for reactive flows , e.g. , flows managed by TCP congestion control algorithms, still remains an open problem. In this paper, we take the first step to solving this problem in a data center network. To address both theoretical and practical challenges, we introduce a novel learning-based prediction system based on the NUM model, with two key techniques named fast factor learning (FFL) and efficient flow sampling . We adopt novel techniques to overcome practical concerns such as scalability, convergence and unknown system parameters. A system, Prophet, is proposed leveraging the emerging technologies of Software Defined Networking (SDN) to realize the model. Evaluations demonstrate that our solution achieves significant accuracy in a wide range of settings.