TCP Congestion Research Articles

Optimization of TCP Congestion Control Algorithm in Dynamic Adaptive Streaming over HTTP

Optimization of TCP Congestion Control Algorithm in Dynamic Adaptive Streaming over HTTP

Janus: A Multi-TCP Framework for Application-Aware Optimization in Mobile Networks

As the dominant protocol on the Internet, TCP has attracted significant attention and has been implemented in various ways, each of which optimizes for a single objective such as high throughput or low delay. However, in today's mobile networks that carry traffic from diverse types of flows, this approach may lead to misconfiguration of TCP congestion control algorithms and further degrade performance for many applications. In this paper, we propose Janus, a new transport-layer framework that automatically selects among existing congestion control variants to optimize traffic in accordance with application demands. Janus is easy to deploy because it reuses existing, well-tested congestion control implementations, and does not require any in-network or client-side changes. To explore the potential for this approach, we implement Janus in the Linux kernel and extensively evaluate its performance with both emulated and real Internet traffic. We show Janus outperforms alternative protocols by offering fast convergence times in response to changing network conditions, achieving 5-10X lower delay with comparable or higher throughput. Our approach also significantly improves user-perceived performance according to QoE metrics, with up to 5X fewer interruptions for video streaming applications and 2X faster page loading for web-browsing applications.

Channel Occupancy Ratio Based TCP Congestion Control in Adhoc Networks

Wireless Network is been used widely in the recent years due to its low-cost nature. More number of real time applications have been used by varied segments of users across the world. The advancements in the mobile technology have made ad hoc networks as important and active field of communication and networks. Most of the time the network cannot handle the traffic in the network which ultimately affects the Quality of Service (QoS). The conventional Transmission Control Protocol (TCP) cannot handle this huge volume of traffic and control the congestion in the network. This issue in the wireless network is been addressed by various researchers in the world but still the scope and need for improvements are more. This paper analyzes many TCP congestion control mechanisms and proposes an efficient approach for congestion control by estimating Channel Occupancy Ratio (COR). The COR is estimated based on machine learning algorithm which is trained using the historical data extracted from MAC layer. This cross layer based approach is found to be more efficient when compared to other methods proposed in the literature.

Performance Improvement of TCP New Reno using one way delay Measurement

Well known TCP NewReno protocol when used with wire-cum-wireless network its performance decreases, due to the inability to discriminate wireless loss with congestion loss. One way delay can give better network characteristics than approximate RTT used in TCP congestion control algorithms. In this paper, we have proposed a modified version of TCP NewReno based on one-way delay measurement. We called it as WNewReno. Our proposed algorithm detects wireless loss and improves performance. Simulations are carried out using ns-2.34.

Extensive evaluation on the performance and behaviour of TCP congestion control protocols under varied network scenarios

In recent decades, many TCP Congestion Control (CC) protocols have been proposed to improve the performance and reliability of TCP in various network scenarios. However, CC protocols are usually closely coupled with network conditions such as latency and packet loss. Considering that networks with different properties are common, e.g., wired/wireless LAN and Long Fat Networks (LFNs), investigating both performance and behaviors of CC protocols under varied network scenarios becomes crucial for both network management and development. In this paper, we conduct a comprehensive measurement study on the goodput, RTT, retransmission, friendliness, fairness, convergence time and stability of most widely-used CC protocols over wired LAN/WAN and wireless LAN (both 2.4GHz and 5GHz Wi-Fi). We also conduct comparative studies with respect to transmission cost, congested reverse path and bottleneck queue size in network simulator.Based on our analysis, we reveal several interesting and original observations. We found that the goodput of BBR is at least 22.5% lower than other CC protocols in wireless LAN due to insufficient pacing rate, even though it can always fully utilize the bottleneck bandwidth with low RTT in wired networks. We also observed that the total on-wire data volume of BBR is higher than CUBIC (e.g., 2.37% higher when RTT = 100ms and loss rate = 0.01%). In addition, BBR can fully utilize the bottleneck bandwidth in most queue sizes ( ≥ 20packets). Surprisingly, we noticed that as the default CC protocol in most modern operating systems, CUBIC is too aggressive and unfriendly in both LAN and wireless LAN, greatly suppressing the goodput of other competing CC protocols. More specifically for CUBIC in wireless LAN, it generates 129% more retransmissions than other CC protocols. Nevertheless, we have also seen that, in scenario with heavily-congested reverse path, CUBIC can provide full utilization on bottleneck bandwidth. Lastly, we also observed that BBR converges very quickly in all evaluated scenarios, while other CC protocols present varied results, e.g., Westwood+ and Veno converge faster in 5GHz Wi-Fi networks than 2.4GHz networks.

Congestion Control in CoAP Observe Group Communication.

The Constrained Application Protocol (CoAP) is a simple and lightweight machine-to-machine (M2M) protocol for constrained devices for use in lossy networks which offers a small memory capacity and limited processing. Designed and developed by the Internet Engineering Task Force (IETF), it functions as an application layer protocol and benefits from reliable delivery and simple congestion control. It is implemented for request/response message exchanges over the User Datagram Protocol (UDP) to support the Internet of Things (IoT). CoAP also provides a basic congestion control mechanism. In dealing with its own congestion, it relies on a fixed interval retransmission timeout (RTO) and binary exponential backoff (BEB). However, the default CoAP congestion control is considered to be unable to effectively perform group communication and observe resources, and it cannot handle rapid, frequent requests. This results in buffer overflow and packet loss. To overcome these problems, we proposed a new congestion control mechanism for CoAP Observe Group Communication, namely Congestion Control Random Early Detection (CoCo-RED), consisting of (1) determining and calculating an RTO timer, (2) a Revised Random Early Detection (RevRED) algorithm which has recently been developed and primarily based on the buffer management of TCP congestion control, and (3) a Fibonacci Pre-Increment Backoff (FPB) algorithm which waits for backoff time prior to retransmission. All the aforementioned algorithms were therefore implemented instead of the default CoAP mechanism. In this study, evaluations were carried out regarding the efficiency of the developed CoCo-RED using a Cooja simulator. The congestion control mechanism can quickly handle the changing behaviors of network communication, and thus it prevents the buffer overflow that leads to congestions. The results of our experiments indicate that CoCo-RED can control congestion more effectively than the default CoAP in every condition.

On SDN-Enabled Online and Dynamic Bandwidth Allocation for Stream Analytics

Data communication in cloud-based distributed stream data analytics often involves a collection of parallel and pipelined TCP flows. As the standard TCP congestion control mechanism and its variants are designed for achieving “fairness” among competing flows and are agnostic to the application layer contexts, the bandwidth allocation among a set of TCP flows traversing bottleneck links often leads to sub-optimal application-layer performance measures, e.g., stream processing throughput or average tuple complete latency. Motivated by this and enabled by the rapid development of the software-defined networking (SDN) techniques, in this paper, we re-investigate the design space of the bandwidth allocation problem and propose a cross-layer framework which utilizes the instantaneous information obtained from the application layer and provides on-the-fly and dynamic bandwidth adjustment algorithms for assisting the stream analytics applications achieving better performance during the runtime. We implement a prototype cross-layer bandwidth allocation framework based on a popular open-source distributed stream processing platform, Apache Storm, together with the OpenDaylight controller, and carry out extensive experiments with real-world analytical workloads on top of a local cluster consisting of ten workstations interconnected by a SDN-enabled fat-tree like testbed. The experiment results clearly validate the effectiveness and efficiency of our proposed framework and algorithms. Finally, we leverage the proposed cross-layer SDN framework and introduce an exemplary mechanism for bandwidth sharing and performance reasoning among multiple active applications and show a case of a point solution on how to approximate application-level fairness.

Comparative analysis of cumulative distribution function for TCP congestion window size and triple‐duplicate period

SummaryVarious version of Transmission Control Protocol (TCP) congestion window model are available in the literature. TCP congestion control is handled in two phases: (a) slow start and (b) congestion avoidance. This paper deals with congestion avoidance phase that is based on additive increase multiplicative decrease (AIMD) mechanism in which window size either increase by one or cuts to half of the previous window size. For distribution of window sizes, a model developed by Yan et al has been presented for two loss indications: (a) triple duplicate (TD) and (b) time‐outs (TOs) only while it does not deal with the window size limitation. It does not allow to cap over the TCP congestion window size if the buffer size is restricted to a fixed value though window size could not move beyond that. The unconstrained window size moves on in the presence of low loss probability. In this paper, we propose a model for the case where the window size is bounded by a maximum value that makes applicable to window size limitation as well. Further, the work has been extended to develop a new model that obtains a cumulative distribution function for TD periods (TDPs). The proposed model is validated on ns‐2, and we conclude that observed results for distribution function are very closed to our proposed model.

TCP’s Initial Window—Deployment in the Wild and Its Impact on Performance

TCP congestion control and particularly its initial congestion window (IW) size is one long-debated topic that can influence Web performance. Its size is, however, assumed to be static by IETF recommendations—despite being network- and application-dependent—and only infrequently changed in its history. To understand if the standardization and research perspective still meets Internet reality, we study the IW configurations in IPv4 and of major content delivery networks (CDNs). We have been regularly inspecting IPv4 for HTTP and TLS servers to investigate their IW configuration and found a steady increase in IETF-recommended configurations. We additionally study how CDNs configure their IWs given their relevance for content distribution. To shed light on network-dependent CDN configurations, we use a globally distributed infrastructure of VPNs giving access to residential access links. We observe that most CDNs are well aware of the IW’s impact and find a high amount of customization that is beyond current Internet standards. We find various initial window configurations, most below 50 segments, yet, with exceptions of up to 100 segments—the tenfold of current standards. This paper highlights that Internet reality has drifted away from recommended and standardized practices. Driven by these findings, we investigate the effects of this new reality on the slow start of Cubic and BBR congestion controlled TCP flows. We find that TCP pacing is a key to enable increased IWs when competing against other traffic.

Experience-Driven Congestion Control: When Multi-Path TCP Meets Deep Reinforcement Learning

In this paper, we aim to study networking problems from a whole new perspective by leveraging emerging deep learning, to develop an experience-driven approach, which enables a network or a protocol to learn the best way to control itself from its own experience (e.g., runtime statistics data), just as a human learns a skill. We present design, implementation and evaluation of a deep reinforcement learning (DRL)-based control framework, DRL-CC (DRL for Congestion Control), which realizes our experience-driven design philosophy on multi-path TCP (MPTCP) congestion control. DRL-CC utilizes a single (instead of multiple independent) agent to dynamically and jointly perform congestion control for all active MPTCP flows on an end host with the objective of maximizing the overall utility. The novelty of our design is to utilize a flexible recurrent neural network, LSTM, under a DRL framework for learning a representation for all active flows and dealing with their dynamics. Moreover, we, for the first time, integrate the above LSTM-based representation network into an actor-critic framework for continuous (congestion) control, which leverages the emerging deterministic policy gradient to train critic, actor, and LSTM networks in an end-to-end manner. We implemented DRL-CC based on the MPTCP implementation in the Linux kernel. The experimental results show that 1) DRL-CC consistently and significantly outperforms a few well-known MPTCP congestion control algorithms in terms of goodput without sacrificing fairness, 2) it is flexible and robust to highly-dynamic network environments with time-varying flows, and 3) it is friendly to regular TCP.

Dynamic TCP Initial Windows and Congestion Control Schemes Through Reinforcement Learning

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.

Reproducible measurements of TCP BBR congestion control

The complexity of evaluating TCP congestion control has increased considerably since its initial development in the 1980s. Several congestion control algorithms following different approaches have been proposed to match the requirements of modern computer networks. We present a framework to analyze different congestion control algorithms using network emulation. The framework is publicly available which provides easy repeatability of our experiments and reproducibility of the results.As a case study, we use our framework to analyze the bottleneck bandwidth and round-trip time (BBR) congestion control algorithm, which was published by Google in 2016. Because of promising initial results, BBR has gained wide-spread attention. As such it has been subject analysis, which showed an increase in performance, but also revealed critical flaws. We verify our framework by reproducing experiments from related work which confirm weaknesses of the current BBR implementation. We also contribute an analysis of BBR’s inter-flow synchronization behavior and its interaction when competing with other congestion control algorithms. Our results show that BBR flows on their own have difficulty to reach a fairness equilibrium and suppress other congestion control algorithms. BBR is still work in progress, thus the framework is useful to validate further updates of BBR by rerunning the given experiments.

A Survey on TCP Congestion Control and its Variants for Wireless Networks

TCP stands for Transmission Control Protocol. It is widely used end-to-end protocol. TCP is high reliability protocol for the wired Network. In general TCP is used for wired Network but it can also be used for both wireless Networks too. TCP works efficiently for the wireless Network too. Algorithms are Slow- Start, Congestion Avoidance, congestion control are used in TCP. This paper presents TCP congestion control mechanisms are like Slow- Start, retransmission and congestion window. Paper also discuss about the reasons for congestion control and TCP variants like TCP Tahoe, TCP Vegas, TCP Reno etc.

Estimating TCP flow completion time distributions

The flow completion time is an important performance metric for network users, and it is used in designing scheduling and congestion control algorithms. The existing TCP flow completion time distributions are calculated from complicated models, and they are only suitable for short TCP flows. This paper is motivated to build a simple model to derive TCP flow completion time distribution that is suitable for both short and long TCP flows. To build such a model, the distribution of the packet transmission latency under TCP congestion control is firstly calculated. The packet transmission latency is the time needed to successfully deliver a packet, including the time for retransmissions due to packet losses. Semi-Markov process is used to model the TCP congestion window evolution. Using the packet transmission latency distribution together with the TCP congestion window size information, a new parallel queueing model which describes how a file is transmitted by a TCP flow is formed. In the parallel queueing model, the flow completion time distribution is derived, and simulation results from NS2 demonstrate the accuracy of the proposed analysis.

Receiver-driven fair congestion control for TCP outcast in data center networks

In the commodity data center networks with multi-rooted tree topologies, the close TCP flows are easily suppressed by the far ones, resulting in large tail flow completion time. The existing solutions, however, are mainly window-based and unable to control the micro-burst congestion, leading to performance degradation. In this paper, we reveal empirically that controlling the micro-burst of highly concurrent flows effectively alleviates this TCP outcast problem under severe congestion. We propose the receiver-driven congestion control OFTCP by coordinating the many-to-one flow transmission. Specifically, OFTCP randomly defers ACK transmissions in a small time granularity to mitigate the aggregate micro-burst and cooperatively adjusts the advertisement windows of each flow based on round trip time to provide fair transmission. We evaluate the performance of OFTCP with Mininet testbed and NS2 simulation under realistic datacenter workloads. Experiment results show that OFTCP achieves throughput fairness and reduces the tail flow completion time by up to 53% compared with state-of-the-art datacenter outcast solutions.

A Survey of Delay-Based and Hybrid TCP Congestion Control Algorithms

Congestion control (CC) has a significant influence on the performance of transmission control protocol (TCP) connections. Over the last three decades, many researchers have extensively studied and proposed a multitude of enhancements to standard TCP CC. However, this topic still inspires both academic and industrial research communities due to the change in Internet application requirements and the evolution of Internet technologies. The standard TCP CC infers network congestion based on packet loss events which leads to long queuing delay when bottleneck buffer size is large. A promising solution to this problem is to use the delay signal (RTT or one-way delay measurements) to infer congestion earlier and react to the congestion before the queuing delay reaches a high value. In this survey paper, we describe the delay signal and the algorithms that completely or partially utilize this type of signal. Additionally, we illustrate standard CC and modern active queue management (AQM) principles and discus the interaction between AQM and the delay signal.

RESEARCH ON TCP/IP CONGESTION CONTROL BASED ON NS-2 NETWORK SIMULATION PLATFORM

The emergence of the Internet facilitates people's lives. Although the network bandwidth increases rapidly, channel resources are limited and network congestion is bound to occur. In order to alleviate network congestion and improve service quality, this paper simply introduced four transmission control protocol (TCP) congestion control algorithms and network simulation-2 (NS-2) network simulation platform for simulating the network environment. Then the performance of four TCP congestion control algorithms in congested network and high bandwidth networks is simulated and analyzed on NS-2 simulation platform. The results showed that TCP Vegas has stable performance in both congested and high-bandwidth networks, and not only the change of cwnd is stable, but also the total number of packets sent and the average throughput are more, and the packet loss rate is 0; the performance of the other three algorithms is consistent in high bandwidth network, In crowded networks, TCP Tahoe performs worst, TCP Reno performs generally, and TCP SACK performs well; TCP Vegas performs poorly on average latency. In summary, TCP Vegas can be used stably in congested networks and high bandwidth networks, but considering the average delay, it is more inclined to apply the other three algorithms in high bandwidth networks.

A Survey on TCP Congestion Control And Its Variants for Wireless Networks

A Survey on TCP Congestion Control And Its Variants for Wireless Networks

Transport layer dependability benchmarking: TCP congestion control with fault injection

Transport layer dependability benchmarking: TCP congestion control with fault injection

Transport layer dependability benchmarking: TCP congestion control with fault injection

Previous approaches have used fault injection to evaluate the distributed system's dependability by targeting the application and hardware layers. In this paper, we delve on the communication layer services which are located between the two previous layers. We provide the specification of the assessment components: fault injection, architectural structure of the workload, and dependability measures. We have examined the use of fault injections to evaluate the dependability of a transport layer protocol (i.e. TCP) with a random study. We focus on the congestion control service, then, we define workload, faultload and dependability measures. We show that the communication is interrupted after exceeding a given tolerance threshold (below this limit, the service employs a recovery time to resume execution). We conclude that congestion control is robust to port fields; however, fault injections to the window size and flag fields hang the service. The test bed is carried out through the simulator NS-3.