TCP Congestion Control Mechanism Research Articles

Utility function of TCP

Understanding the TCP congestion control mechanism from a global optimization point of view is not only important in its own right, but also crucial to the design of other transport layer traffic control protocols with provable properties. In this paper, we derive a global utility function and the corresponding optimal control law, known as TCP control law, which maximizes the global utility. The TCP control law captures the essential behaviors of TCP, including slow start, congestion avoidance, and the binary nature of congestion feedback in TCP. We find that the utility function of TCP is linear in the slow start phase and is proportional to the additive increase rate and approaches the well-known logarithm function as the data rate becomes large in the congestion avoidance phase. We also find that understanding the slow start phase with a fixed threshold is critical to the design of new transport layer control protocols to enable quality of service features. Finally, as an application, we design a Minimum Rate Guaranteed (MRG) traffic control law that shares the same utility function as the TCP control law. Our simulation study of the MRG control law indicates that it is indeed TCP friendly and can provide minimum rate guarantee as long as the percentage of network resource consumed by the MRG flows is moderately small.

TCP Congestion Control Mechanisms for Achieving Predictable Throughput Using Inline Network Measurement

Recently, real-time media delivery services such as video streaming and VoIP have rapidly become popular. For these applications requiring high-level QoS guarantee, our research group has proposed a transport-layer approach to provide predictable throughput for upper-layer applications. In the present paper, we propose a congestion control mechanism of TCP for achieving predictable throughput. It does not mean we can guarantee the throughput, while we can provide the throughput required by an upper-layer application at high probability when network congestion level is not so high by using the inline network measurement technique for available bandwidth of the network path. We present the evaluation results for the proposed mechanism obtained in simulation and implementation experiments, and confirm that the proposed mechanism can assure a TCP throughput if the required bandwidth is not so high compared to the physical bandwidth, even when other ordinary TCP (e.g., TCP Reno) connections occupy the link.

On TCP performance enhancing proxies in a wireless environment

TCP performance enhancement in wireless access networks is an important ongoing area of research. It is known that the hostile nature of the wireless channel and the mobile nature of wireless users interact adversely with standard TCP congestion control mechanisms [1], causing a drastic reduction in throughput. This article surveys a selection of different approaches to managing TCP performance over wireless links, and presents the results of simulation and field trial results of a novel TCP performance enhancing proxy over diverse cellular radio access technologies based on the GSM, cdma2000, and UMTS standards. The proposed TRL TCP performance enhancing proxy has the advantages of being completely transparent to both TCP endpoints and tunable to different access technologies, and frequently achieves the maximum throughput available from any of the studied radio access technologies.

Simulation comparison of active queue management algorithms in TCP/IP networks

Active Queue Management techniques are recommended to overcome the performance limitations of TCP congestion control mechanisms over drop-tail networks. The main objective of this paper is to present the comparative analysis of the performance of 10 different queue management policies using the OMNeT++ simulator. The AQMs have many parameters which influence their behavior, so some results concerning the parameters selection are given. The algorithms are tested in terms of average queue size, packet delay and packet loss rate in presence of TCP traffic.

TCP-Cherry: A new approach for TCP congestion control over satellite IP networks

TCP performance is vulnerable to long delays and errors over satellite links despite their potential for ubiquitous coverage. In this paper, we therefore propose a new TCP congestion control mechanism, TCP-Cherry, to improve TCP performance over satellite IP networks. The basic idea is to probe the network for available resources using data segments that do not put the network into congestion. To this goal, TCP-Cherry deploys a new type of low-priority data segments, namely, supplement segments that in addition to probing the network, carry data not yet transmitted as regular data segments. Our major contributions in this work include two new algorithms, Fast-Forward Start and First-Aid Recovery, for congestion control in the presence of data segments belonging to two different levels of priority. Another unique feature of our scheme is the mechanism for selection of supplement segments which keeps the overhead of duplicate transmission minimum. We also propose a simple scheme for avoiding gradual decrease of TCP congestion window owing to recurring loss of probing segments. Simulation results show that our proposed TCP-Cherry yields upto a maximum improvement of more than 150% in goodput compared with other existing schemes. We also observe that TCP-Cherry maintains a fair share of bandwidth among competing TCP connections. In addition, the supplement segments used for probing the network induce the minimal network overhead.

Performance Analysis of Large-Scale IP Networks Considering TCP Traffic

In this paper, we propose a novel analysis method for large-scale networks with consideration of the behavior of the congestion control mechanism of TCP. In the analysis, we model the behavior of TCP at end-host and network link as independent systems, and combine them into a single system in order to analyze the entire network. Using this analysis, we can analyze a large-scale network, i.e. with over 100/1,000/10,000 routers/hosts/links and 100,000 TCP connections very rapidly. Especially, a calculation time of our analysis, it is different from that of ns-2, is independent of a network bandwidth and/or propagation delay. Specifically, we can derive the utilization of the network links, the packet loss ratio of the link buffer, the round-trip time (RTT) and the throughput of TCP connections, and the location and degree of the network congestion. We validate our approximate analysis by comparing analytic results with simulation ones. We also show that our analysis method treats the behavior of TCP connection in a large-scale network appropriately.

Cross‐layer design to improve elastic traffic performance in WLANs

AbstractIn this paper, we are interested in improving TCP flow performance when a short loss of 802.11 signal leads to losing segments and triggers inappropriately TCP congestion control mechanisms. A set of measurements in a common wireless environment with signal losses due to mobility or interference is made to highlight the distinct MAC and TCP loss recovery levels and the lack of interactions between them. Initially, we demonstrate the interest of adapting the 802.11 MAC layer Retry Limit parameter in the case of signal losses due to distance or obstacles (mobility). Thus, a first‐level loss differentiation algorithm (LDA) acting at the MAC layer is proposed to improve TCP flow performance in the case of segment losses due to mobility. Hence, for a signal failure, the MAC layer reacts consequently by dynamically adapting the Retry Limit parameter. This adaptation allows avoiding a costly end‐to‐end TCP loss recovery. Segment losses due to interference are differentiated from those due to congestion through the use of a second‐level LDA. The latter is a cross‐layer LDA acting at the TCP layer but using a specific 802.11 parameter, the AckFailureCount, to realize the targeted loss differentiation. The TCP NewReno version is then adapted in order to integrate the cross‐layer LDA results and to avoid reducing the TCP congestion window unsuitably. The efficiency and completeness of a solution integrating both LDA schemes is then discussed. Copyright © 2007 John Wiley & Sons, Ltd.

Improving Transport Layer Performance in Multihop Ad Hoc Networks by Exploiting MAC Layer Information

The traditional TCP congestion control mechanism encounters a number of new problems and suffers a poor performance when the IEEE 802.11 MAC protocol is used in multihop ad hoc networks. Many of the problems result from medium contention at the MAC layer. In this paper, we first illustrate that severe medium contention and congestion are intimately coupled, and TCP's congestion control algorithm becomes too coarse in its granularity, causing throughput instability and excessively long delay. Further, we illustrate TCP's severe unfairness problem due to the medium contention and the tradeoff between aggregate throughput and fairness. Then, based on the novel use of channel busyness ratio, a more accurate metric to characterize the network utilization and congestion status, we propose a new wireless congestion control protocol (WCCP) to efficiently and fairly support the transport service in multihop ad hoc networks. In this protocol, each forwarding node along a traffic flow exercises the inter-node and intra-node fair resource allocation and determines the MAC layer feedback accordingly. The end-to-end feedback, which is ultimately determined by the bottleneck node along the flow, is carried back to the source to control its sending rate. Extensive simulations show that WCCP significantly outperforms traditional TCP in terms of channel utilization, delay, and fairness, and eliminates the starvation problem

Extending TCP congestion control to multicast

This paper proposes a new single-rate multicast congestion control scheme named PGMTCC, which has been implemented and investigated in PGM. The primary idea of PGMTCC is to extend Sack TCP congestion control mechanism to multicast in order to make multicast perform almost the same as Sack TCP under all kinds of network conditions. To achieve this goal, first of all, the sender should accurately select a receiver with the worst throughput as a representative (acker) by a simplified equation of TCP throughput. Then the Sack TCP congestion control mechanism, with some modifications to be adapted to multicast, is deployed to take charge of congestion control between the sender and the acker. Moreover, in our scheme, the problem of the feedback suppression is considered and solved by a selective suppression mechanism of feedback. NS2 is used to test and investigate the performance of our scheme. As expected, PGMTCC performs almost like Sack TCP under all kinds of conditions. We believe that it is TCP-friendly, robust and scalable.

Modelling Internet End-to-End Loss Behaviors: A Congestion Control Perspective

This paper proposes a new approach to modelling and controlling Internet end-to-end loss behaviours. Rather than select the model structure from the loss observations as being done previously, we construct a new loss model based on the TCP congestion control mechanisms. Thus, the model can explicitly reflect the correlation between end-to-end loss observations and network flow level activities. Besides simulation, the model has been tested in both wired and wireless Internet environments. The result shows that, unless the losses due to the transmission errors are excessive e.g. in some lossy wireless channels, the model can correctly capture end-to-end loss behaviours not only in terms of average rates but also in terms of loss patterns i.e. loss and good runlengths. This implies a good connection between the model structure and network flow level activities, which makes the model attractive for assisting network traffic control.

Performance evaluation of high speed TCP over optical burst switching networks

Study of TCP performance over OBS networks has been an important problem of research lately and it was found that due to the congestion control mechanism of TCP and the inherent bursty losses in the Optical Burst Switching (OBS) network, the throughput of TCP connections degrade. On the other hand, High Speed TCP (HSTCP) was proposed as an alternative to the use of TCP in high bandwidth-delay product networks. HSTCP aggressively increases the congestion window used in TCP, when the available bandwidth is high and decreases the window cautiously in response to a congestion event. In this work, we make a thorough simulation study of HSTCP over OBS networks. While the earlier works in the literature used a linear chain of nodes as the network topology for the simulation, we use the popular 14-node NSFNET topology that represents an arbitrary mesh network in our study. We also study the performance of HSTCP over OBS for different bandwidths of access networks. We use two different cases for simulations where in the first HSTCP connections are routed on disjoint paths while in the second they contend for resources in the network links. These cases of simulations along with the mesh topology help us clearly distinguish between the congestion and contention losses in the OBS network and their effect on HSTCP throughput. For completeness of study, we also simulate TCP traffic over OBS networks in all these cases and compare its throughput with that of HSTCP. We observe that irrespective of the access network bandwidth and the burst loss rate in the network, HSTCP outperforms TCP in terms of the throughput and robustness against multiple burst losses up to the expected theoretical burst loss probability of 10 −3.

Virtual source-based minimum interference path multicast routing in optical virtual private networks

A major concern in optical burst-switched (OBS) networks is contention, which occurs when more than one bursts contend for the same data channel at the same time. Due to the bufferless nature of OBS networks, these contentions randomly occur at any degree of congestion in the network. When contention occurs at any core node, the core node drops bursts according to its dropping policy. Burst loss in OBS networks significantly degrades the throughput of TCP sources in the local access networks because current TCP congestion control mechanisms perform a slow start phase mainly due to contention rather than heavy congestion. However, there has not been much study about the impact of burst loss on the performance of TCP over OBS networks. To improve TCP throughput over OBS networks, we first introduce a dropping policy with burst retransmission that retransmits the bursts dropped due to contention, at the ingress node. Then, we extend the dropping policy with burst retransmission to drop a burst that has experienced fewer retransmissions in the event of contention at a core node in order to reduce the number of events that a TCP source enters the slow start phase due to contention. In addition, we propose to limit the number of retransmissions of each burst to prevent severe congestion. For the performance evaluation of the proposed schemes, we provide an analytic throughput model of TCP over OBS networks. Through simulations as well as analytic modeling, it is shown that the proposed dropping policy with burst retransmission can improve TCP throughput over OBS networks compared with an existing dropping policy without burst retransmission.

Influence of TCP congestion control on multifractal nature of network traffic

AbstractThe multifractal nature of traffic volume is known to have significant influence to queuing behavior in network nodes. In the Internet, TCP traffic, which is the major traffic flowing through the Internet has been reported to have multifractal nature. To control the effect of the multifractal nature, the origin of the multifractal nature in TCP traffic needs to be explained. However, it is still unclear what part of TCP control mostly affects the multifractal nature. In this paper, we focus on TCP congestion control from the viewpoint of multifractal nature of generated traffic. We investigate “what component of TCP congestion control mechanism affects multifractal nature of generated traffic?” Our simulation results show that Retransmission Time‐Out mechanism plays an important role for multifractal nature of TCP traffic. In addition, our results show that avoiding occurrence of Retransmission Time Out is the best approach to improve TCP performance, i.e., throughput degradation, LRD, and multifractal nature. Random Early Detection (RED), based on the fact that Retransmission Time Out should be avoided, to improve TCP performance are also reasonable from the viewpoint of LRD and multifractal nature. © 2006 Wiley Periodicals, Inc. Electron Comm Jpn Pt 1, 90(1): 68–76, 2007; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecja.20277

Impact of Bluetooth MAC layer on the performance of TCP traffic

AbstractRecent updates of the Bluetooth specification have introduced significant changes in the Bluetooth protocol stack, including optional flow control. When the Bluetooth piconet is used to carry TCP traffic, complex interactions between TCP congestion control mechanisms and data link layer controls of Bluetooth will occur. In this paper, we model the performance of the piconet with TCP traffic, expressed through segment loss probability, round‐trip time, and goodput, through both probabilistic analysis and discrete‐event simulations. We show that satisfactory performance for TCP traffic may be obtained through proper dimensioning of the Bluetooth architecture parameters. Copyright © 2006 John Wiley & Sons, Ltd.

TCP BaLDE for Improving TCP Performance over Heterogeneous Networks

Network congestion and random errors of wireless link are two well-known noteworthy parameters which degrade the TCP performance over heterogeneous networks. We put forward a novel end-to-end TCP congestion control mechanism, namely TCP BaLDE (Bandwidth and Loss Differentiation Estimate), in which the TCP congestion control categorizes the reason of the packet loss by estimating loss differentiation in order to control the packet transmission rate appropriately. While controlling transmission rate depends on the available bandwidth estimation which is apprehended by the bandwidth estimation algorithm when the sender receives a new ACK with incipient congestion signal, duplicates ACKs or is triggered by retransmission timeout event. Especially, this helps the sender to avoid router queue overflow by opportunely entering the congestion avoidance phase. In simulation, we experimented under numerous different network conditions. The results show that TCP BaLDE can achieve robustness in aspect of stability, accuracy and rapidity of the estimate in comparison with TCP Westwood, and tolerate ACK compression. It can achieve better performance than TCP Reno and TCP Westwood. Moreover, it is fair on bottleneck sharing to multiple TCP flows of the same TCP version, and friendly to existing TCP version.

Cross-Layer Design Improves TCP Performance in Multihop Ad Hoc Networks

TCP performance in the IEEE 802.11-based multihop ad hoc networks is extremely poor, because the congestion control mechanism of TCP cannot effectively deal with the problem of packet drops caused by mobility and shared channel contention among wireless nodes. In this paper, we present a cross-layer method, which adaptively adjusts the TCP maximum window size according to the number of RTS (Request To Send) retry counts of the MAC layer at the TCP sender, to control the number of TCP packets in the network and thus decrease the channel contention. Our simulation results show that this method can remarkably improve TCP throughput and its stability.

Congestion control based on flow-state-dependent dynamic priority scheduling

The existing congestion control mechanism in the current Internet use is based on TCP congestion control mechanism, which is the key to the success of the Internet. However the TCP's end-to-end congestion control mechanism and the router assistant mechanism active queue management (AQM), such as random early detection (RED), cannot cope with flows that are unresponsive to congestion notification or non-TCP-compatible flows which are responsive but more aggressive than TCP. Unresponsive or non-TCP compatible flows could even cause congestion collapse. In this paper, a new mechanism is presented, namely source aware congestion control (SACC) based on a packet scheduling algorithm called flow-state-dependent dynamic priority scheduling (FDPS), which performs traffic monitoring, marking and scheduling at routers. Extensive simulation experiments are conducted based on ns-2 to examine SACC behaviour and evaluate its performance in a variety of scenarios. The numerical results illustrate that the proposed scheme can differentiate services and resolve congestion by gracefully discarding packets from non-conformant and best effort flows when unresponsive flows exist.

De l'utilisation des mesures de trafic pour l'ingénierie des réseaux de l'internet

Monitoring is becoming the cornerstone of network engineering and research, in particular for the Internet. Because of the increasing complexity of the Internet, the traffic characteristics and network usages are badly known. In such conditions, it become harder and harder to fully control the network and its traffic. We propose in this paper a new based methodology that integrates in all its stages some measurements. First stage consists in characterizing the network traffic and its dynamics. The second consists in using monitoring to improve the realism of simulations, in order to evaluate architectures and protocols in realistic environments. First results on traffic characterization show an oscillatory phenomenon, at all scales, that is a major issue for QoS. These oscillations seem to be associated to some long range dependence feature (LRD) in the traffic. This paper then studies possible causes of LRD and how LRD can impact QoS. In particular TCP congestion control mechanisms are deeply analyzed. This paper then proposes to illustrate our based methodology. For that, it is proposed to modify some TCP features to reduce LRD and then improve network QoS. MOTS-CLES : metrologie Internet, caracterisation du trafic, optimisation, qualite de service, controle de congestion, TCP, TFRC.

Error differentiation with measurements based on wave patterns

We investigate the efficiency and fairness potential of an error differentiation mechanism with measurements based on wave patterns. Our work is focused on a receiver-oriented protocol, TCP-Real that allows for measurement-based congestion detection and error recovery instead of a ‘blind’ increase/decrease window adjustment. We conduct a comparative evaluation of congestion control mechanisms of standard TCP, TCP-Real, and TCP-friendly protocols in multiplexed wired/wireless networks for both delay-tolerant and delay-sensitive applications. We show that the receiver-oriented communication of TCP-Real (i) diminishes the impact of reverse path asymmetry on the transmission rate, (ii) allows for error recovery tactics that are responsive to the nature of the error detected (i.e. congestion versus bit corruption), and (iii) dictates transmission rate adjustments that conform to the level of present contention. As a direct impact of this design, we realize the inherent properties of our protocol to produce comprehensive dynamics in heterogeneous environments, in contrast to the occasionally limited efficiency of standard TCP and the application-specific design of TCP-friendly protocols.

Transport of TCP/IP traffic over assured forwarding IP-differentiated services

The Internet is facing a twofold challenge: to increase network capacity in order to accommodate a steadily increasing number of users; to guarantee the quality of service for existing applications and for new multimedia applications requiring real-time network response. In order to meet these requirements, IETF is currently defining the differentiated service (DiffServ) architecture, which should offer a simple and scalable platform to guarantee differentiated QoS in the Internet. In the DiffServ domain, the assured forwarding service is designed to provide data applications with acceptable performance, overcoming the limits of the Internet's current best-effort service. Since data applications mostly rely on the TCP transport protocol, it is important to examine the interaction between the congestion avoidance and control mechanisms of TCP and assured forwarding. Our main purpose is to shed light on this interaction, and to show that, in the current DiffServ framework, poor performance of TCP traffic flows can result from the existing mismatch between the assured forwarding traffic conditioning procedures and the TCP congestion management. We propose a new adaptive packet marking policy to deal with congestion situations that may occur. We show that, with this policy, the provisioned rate for TCP flows can be achieved.