TCP Libra: Derivation,analysis, and comparison with other RTT-fair TCPs

The Transmission Control Protocol (TCP), the most widely used transport protocol over the Internet, has been advertised to implement fairness between flows competing for the same narrow link. However, when session round-trip-times (RTTs) radically differ, the share may be anything but fair. This RTT-unfairness represents a problem that severely affects the performance of long-RTT flows and whose solution requires a revision of TCP’s congestion control scheme. To this aim, we discuss TCP Libra, a new transport protocol able to ensure fairness and scalability regardless of the RTT, while remaining friendly towards legacy TCP. As main contributions of this paper: (i) we focus on the model derivation and show how it leads to the design of TCP Libra; (ii) we analyze the role of its parameters and suggest how they may be adjusted to lead to asymptotic stability and fast convergence; (iii) we perform model-based, simulative, and real testbed comparisons with other TCP versions that have been reported as RTT-fair in the literature. Results demonstrate the ability of TCP Libra in ensuring RTT-fairness while remaining throughput efficient and friendly towards legacy TCP.     

一种基于H-TCP的拥塞控制改进算法

针对H-TCP存在的RTT公平性和TCP友好性不好的缺点,提出了一种基于H-TCP的拥塞控制改进算法RH-TCP。NS-2模拟实验结果表明,RH-TCP明显提高了RTT公平性和TCP友好性,并且保留了良好的带宽利用率、稳定性、收敛性和公平性,在6个新TCP协议中RH-TCP的总体性能最好。     

拥塞控制端算法的实验研究

利用网络模拟软件NS-2对最近提出的端算法HSTCP、STCP、BIC-TCP、H-TCP和Fast-TCP进行了实验研究,从收敛性、稳定性、协议内公平性、RTT公平性、TCP友好性和带宽利用率等方面来比较它们的性能.     

Adaptive congestion protocol: A congestion control protocol with learning capability

There is strong evidence that the current implementation of TCP will perform poorly in future high-speed networks. To address this problem many congestion control protocols have been proposed in literature which, however, fail to satisfy key design requirements of congestion control protocols, as these are outlined in the paper. In this work we develop an adaptive congestion protocol (ACP) which is shown to satisfy all the design requirements and thus outperform previous proposals. Extensive simulations indicate that the protocol is able to guide the network to a stable equilibrium which is characterized by max–min fairness, high-utilization, small queue sizes and no observable packet drops. In addition, it is found to be scalable with respect to changing bandwidths, delays and number of users utilizing the network. The protocol also exhibits nice transient properties such as smooth responses with no oscillations and fast convergence. In realistic traffic scenarios comprising of a small number of long flows and a large number of short flows, ACP outperforms both TCP and XCP, even in the presence of random packet losses. ACP does not require maintenance of per flow states within the network and utilizes an explicit multi-bit feedback signalling scheme. To maintain stability it implements at each link a novel estimation algorithm which estimates the number of flows utilizing the link. Using a simple network model, we show analytically the effectiveness of the estimation algorithm. We use the same model to generate phase portraits which demonstrate that the ACP protocol is stable for all delays.     

Use of competition detection in TCP for simple topology networks

The fairness (or TCP-friendliness) of recent high-speed TCP proposals for high bandwidth-delay product networks is generally poor. We believe that the lack of TCP-friendliness of high-speed TCP proposals stems from their ineffectiveness in detecting competing TCP flows. We suggest a competition detection mechanism for a single TCP flow to detect the presence of competing TCP flows. We propose a new TCP, called Adaptive TCP (A-TCP) to demonstrate the usefulness of the competition detection mechanism. A-TCP uses the competition detection mechanism to control its aggressiveness: If it does not detect competing flows, a single A-TCP flow increases its sending rate aggressively in order to highly utilize the network. Otherwise, it behaves like a standard TCP flow to fairly share network resources with competing flows. We implemented A-TCP as part of Linux as well as in ns-2. Experimental results show that A-TCP achieves better fairness than existing high-speed TCP proposals when they compete against standard TCP in simple topology networks.     

Unreliable transport protocol using congestion control for high-speed networks

Currently there is no control for the real-time traffic of multimedia applications using UDP (User Datagram Protocol) in high-speed networks. Therefore, although a number of high-speed TCP (Transmission Control Protocol) protocols have been developed for gigabit-speed (or faster) links, the real-time traffic could also congest the network and result in unfairness and throughput degradation of TCP traffic. In this paper, a new unreliable transport protocol, FAST DCCP, is presented for the real-time traffic in high-speed networks. FAST DCCP is based on the DCCP protocol and adopts the FAST scheme to realize congestion control. Some modifications have been made to the mechanisms inherited from DCCP so as to let the proposed protocol can efficiently operate under a large size window. In addition, an enhanced protocol, EEFAST DCCP, using the measurements of one-way delay to dynamically adjust the window size is also proposed to improve the throughput of FAST DCCP with the effect of reverse traffic. Simulation results show that FAST DCCP not only can satisfy the requirements of real-time data delivery, but also perform well in bandwidth utilization and fairness in high-speed networks. Meanwhile, EEFAST DCCP is able to effectively conquer the throughput degradation caused by the reverse traffic.     

Can high-speed networks survive with DropTail queues management?

It has been observed that TCP connections that go through multiple congested links (MCL) have a smaller transmission rate than the other connections. Such TCP behavior is a result of two components (i) the cumulative packet losses that a flow experiences at each router along its path; (ii) the longer round trip times (RTTs) suffered by such flows due to non-negligible queueing delays at congested routers. This double “bias” against connections with MCLs has been shown to approximate the so-called minimum potential delay fairness principle in the current Internet. Despite the recent proliferation of new congestion control proposals for TCP in high-speed networks, it is still unclear what kind of fairness principle could be achieved with such newly proposed congestion control protocols in high-speed networks with large-delays. Studies already show that some high-speed TCP variants may cause surprisingly severe RTT unfairness in high-speed networks with DropTail routers.This paper studies the problem of unfairness in high-speed networks with some well-known high-speed TCP variants in presence of multiple congested links and highlights the severity of such unfairness when DropTail queue management is adopted.Through a simple synchronized loss model analysis, we show how synchronized losses with DropTail in high-speed networks could lead to severe RTT unfairness and drop probability (DP) unfairness; while random marking AQM schemes, which break the packet loss synchrony mitigate such unfairness dramatically by ensuring that the packet loss probability of a flow is the sum of the loss probabilities on the congested routers it crosses.Extensive simulations are carried out and the results support our findings.     

基于仿真的高速长距离网络中TCP协议性能评价

新的传输控制协议不断涌现,迫切需要一个系统的评价体系。提出了一个系统的和可重复的评价方法,定义了一系列性能测度集和仿真用例,基于NS2对最近几年提出的几个具有代表性的高速长距离网络中TCP协议High-Speed TCP、Scalable-TCP、FAST-TCP、BIC-TCP以及H—TCP的性能进行了比较和评价。仿真结果表明,通过一致的应用性能测度和标准仿真用例可以获得有价值的结论。     

Evaluation of Advanced TCP Stacks on Fast Long-Distance Production Networks

With the growing needs of data intensive science, such as high energy physics, and the need to share data between multiple remote computer and data centers worldwide, the necessity for high network performance to replicate large volumes (TBytes) of data between remote sites in Europe, Japan and the U.S. is imperative. Currently, most production bulk-data replication on the network utilizes multiple parallel standard (Reno based) TCP streams. Optimizing the window sizes and number of parallel stream is time consuming, complex, and varies (in some cases hour by hour) depending on network configurations and loads. We therefore evaluated new advanced TCP stacks that do not require multiple parallel streams while giving good performances on high speed long-distance network paths. In this paper, we report measurements made on real production networks with various TCP implementations on paths with different Round Trip Times (RTT) using both optimal and sub-optimal window sizes.We compared the New Reno TCP with the following stacks: HS-TCP, Fast TCP, S-TCP, HSTCP-LP, H-TCP and Bic-TCP. The analysis will compare and report on the stacks in terms of achievable throughput, impact on RTT, intra- and inter-protocol fairness, stability, as well as the impact of reverse traffic.We also report on some tentative results from tests made on unloaded 10 Gbps paths during SuperComputing 2003.     

Logarithmic window increase for TCP Westwood+ for improvement in high speed,long distance networks

The majority of current Internet applications uses Transmission Control Protocol (TCP) for ensuring reliable end-to-end delivery of data over IP networks. The resulting path is, generally speaking, characterized by fairly large propagation delays (of the order of tens to hundreds of milliseconds) and increasing available bandwidth. Current TCP¹performance is far from representing an optimal solution in such operating conditions. The main reason lies in the conservative congestion control strategy employed, which does not let TCP to exploit the always increasing available path capacity. As a consequence, TCP optimization has been an active research topic in the research community over the last 25 years, boosted in the last few years by the widespread adoption of high-speed optical fiber links in the backbone and the emergence of supercomputing networked applications from one side and tremendous growth of wireless bandwidth in network access from another. This has led to the introduction of several alternative proposals for performing congestion control. Most of them focus on the effectiveness of bandwidth utilization, introducing more “aggressive” congestion control strategies. However, such approaches result often in unfairness among flows with substantially different RTTs, or do not present the inter-protocol fairness features required for incremental network deployment.In this paper, we propose TCP LogWestwood+, a TCP Westwood+ enhancement based on a logarithmic increase function, targeting adaptation to the high-speed wireless environment. The algorithm shows low sensitivity with respect to RTT value, while maintaining high network utilization in a wide range of network settings. The performance, fairness and stability properties of the proposed TCP LogWestwood+ are studied analytically, and then validated by means of an extensive set of experiments including computer simulations and wide area Internet measurements.     

An Improved Link Model for Window Flow Control and Its Application to FAST TCP

This paper presents a link model which captures the queue dynamics in response to a change in a transmission control protocol (TCP) source's congestion window. By considering both self-clocking and the link integrator effect, the model generalizes existing models and is shown to be more accurate by both open loop and closed loop packet level simulations. It reduces to the known static link model when flows' round trip delays are identical, and approximates the standard integrator link model when there is significant cross traffic. We apply this model to the stability analysis of fast active queue management scalable TCP (FAST TCP) including its filter dynamics. Under this model, the FAST control law is linearly stable for a single bottleneck link with an arbitrary distribution of round trip delays. This result resolves the notable discrepancy between empirical observations and previous theoretical predictions. The analysis highlights the critical role of self-clocking in TCP stability, and the proof technique is new and less conservative than existing ones.      

Investigating the performance of TCP in mobile ad hoc networks

Mobile ad hoc networks have several inherent characteristics (e.g. dynamic topology, time-varying and bandwidth constrained wireless channels, multi-hop routing, and distributed control and management). The goal of this work is to investigate the impact of these characteristics on the performance of TCP. First, we investigate throughput performance of TCP as a function of path length (i.e. multiple wireless hops), node mobility, and traffic intensity. Next, we examine the ‘fairness’ of the ad hoc network with regard to equal sharing of network bandwidth among multiple TCP flows. Third, we evaluate the impact of two on-demand routing protocols (i.e. AODV and DSR) on the throughput of TCP. Finally, a factorial design experiment is conducted to quantify the effects and interactions of three factors, which influence the throughput of TCP. These factors include routing, node speed, and node pause time. Two key results were observed. Results show that traffic intensity (e.g. number of concurrent flows) is significantly affects TCP throughput, suggesting the need for congestion control, scheduling and traffic management schemes. Second, source routing achieves higher throughputs while also generating significantly less routing overhead than AODV. Results also show that in some instances, the fairness of the network is very uneven among concurrent TCP flows, resulting in several sending stations achieving very little or no throughput.     

一种基于HSTCP改进的公平性算法

在高速网络中现有的标准TCP不能充分的利用网络带宽,HSTCP(HighSpeed TCP)作为解决这一问题的可行方法被提出.试验表明HSTCP比标准TCP能够更充分的利用带宽,但存在着严重的RTT不公平性.首先通过仿真试验和数学分析对HSTCP的RTT不公平性进行研究,然后在原有算法的基础上添加一个公平性因子来降低由于RTT不同造成的窗口增长差异.试验表明改进算法有效保证了HSTCP流的带宽公平性、降低了丢包率.     

高速TCP/AQM中的RED算法分析

通过建立适用于高速TCP和AQM反馈控制系统的流体流模型,分析高速TCP/AQM闭环系统的稳定性。采用频域稳定裕度,得到高速TCP/AQM中RED算法的稳定参数区域。基于MATLAB/SIMULINK的仿真结果验证了该方法的有效性。     

CSFQ算法分析与改进

核心无状态公平队列调度（CSVQ）算法提供了如同有状态网那样好的公平带宽分配,但它的丢包算法不适用于TCP流。针对TCP流的特点,对CSFQ算法进行如下改进：将缓存队列长度与丢包概率关联起来,用一种类似于RED（random early drop）缓存管理方法解决了缓存频繁溢出导致的一些问题;对TCP流的丢包率进行修正,使用多余带宽来转发TCP包,解决TCP流与UDP流的带宽分配公平性。仿真试验表明,新算法NEW-CSFQ更好地提供数据流公平的频宽共享,对突发流响应较原算法有所提高,且算法复杂度简单,容易在高速核心路由器上实现。     

TCP with gateway adaptive pacing for multihop wireless networks with Internet connectivity

This paper introduces an effective congestion control pacing scheme for TCP over multihop wireless networks with Internet connectivity. The pacing scheme is implemented at the wireless TCP sender as well as at the Internet gateway, and reacts according to the direction of TCP flows running across the wireless network and the Internet. Moreover, we analyze the causes for the unfairness of oncoming TCP flows and propose a scheme to throttle aggressive wired-to-wireless TCP flows at the Internet gateway to achieve nearly optimal fairness. The proposed scheme, which we denote as TCP with Gateway Adaptive Pacing (TCP-GAP), does not impose any control traffic overhead for achieving fairness among active TCP flows and can be incrementally deployed since it does not require any modifications of TCP in the wired part of the network. In an extensive set of experiments using ns-2 we show that TCP-GAP is highly responsive to varying traffic conditions, provides nearly optimal fairness in all scenarios and achieves up to 42% more goodput for FTP-like traffic as well as up to 70% more goodput for HTTP-like traffic than TCP NewReno. We also investigate the sensitivity of the considered TCP variants to different bandwidths of the wired and wireless links with respect to both aggregate goodput and fairness.     

Evaluating TCP-friendliness in light of Concurrent Multipath Transfer

In prior work, a CMT protocol using SCTP multihoming (termed SCTP-based CMT) was proposed and investigated for improving application throughput. SCTP-based CMT was studied in (bottleneck-independent) wired networking scenarios with ns-2 simulations. This paper studies the TCP-friendliness of CMT in the Internet. In this paper, we surveyed historical developments of the TCP-friendliness concept and argued that the original TCP-friendliness doctrine should be extended to incorporate multihoming and SCTP-based CMT.Since CMT is based on (single-homed) SCTP, we first investigated TCP-friendliness of single-homed SCTP. We discovered that although SCTP’s congestion control mechanisms were intended to be “similar” to TCP’s, being a newer protocol, SCTP specification has some of the proposed TCP enhancements already incorporated which results in SCTP performing better than TCP. Therefore, SCTP obtains larger share of the bandwidth when competing with a TCP flavor that does not have similar enhancements. We concluded that SCTP is TCP-friendly, but achieves higher throughput than TCP, due to SCTP’s better loss recovery mechanisms just as TCP-SACK and TCP-Reno perform better than TCP-Tahoe.We then investigated the TCP-friendliness of CMT. Via QualNet simulations, we found out that one two-homed CMT association has similar or worse performance (for smaller number of competing TCP flows) than the aggregated performance of two independent, single-homed SCTP associations while sharing the link with other TCP connections, for the reason that a CMT flow creates a burstier data traffic than independent SCTP flows. When compared to the aggregated performance of two-independent TCP connections, one two-homed CMT obtains a higher share of the tight link bandwidth because of better loss recovery mechanisms in CMT. In addition, sharing of ACK information makes CMT more resilient to losses. Although CMT obtains higher throughput than two independent TCP flows, CMT’s AIMD-based congestion control mechanism allows other TCP flows to co-exist in the network. Therefore, we concluded that CMT is TCP-friendly, similar to two TCP-Reno flows are TCP-friendly when compared to two TCP-Tahoe flows.     

On Sustained QoS Guarantees in Operated IEEE 802.11 Wireless LANs

Most of QoS-capable IEEE 802.11 MAC protocols are unable to deliver sustained quality of service while maintaining high network utilization, particularly under congested network conditions. The problem often resides in the fact that flows belonging to the same service class are assigned the same MAC parameters regardless theirs respective bitrate, which leads to throughput fairness rather than perceived QoS fairness. Harmonizing MAC parameters of traffic classes's flows may further lead to sub-optimal situations since certain network configurations (in terms of per class traffic load) can not be accommodated without reassigning the basic MAC parameters. In this paper, we propose a new cross-layer MAC design featuring a delay-sensitive backoff range adaptation along with a distributed flow admission control. By monitoring both MAC queue dynamics and network conditions, each traffic class reacts based on the degree to which application QoS metrics (delay) are satisfied. Besides, we use a distributed admission control mechanism to accept new flows while protecting the active one. Simulation results show that compared to the enhanced distributed coordination function (EDCA) scheme of 802.11e, our protocol consistently excels, in terms of network utilization, bounded delays, and service-level fairness.     

RAAR:多媒体流在Internet上传输的一种TCP-Friendly拥塞控制机制

提出了一种接收端的速率自适应算法，称为RAAR。它可以应用于单播传输多媒体数据业务。TCP由于它的遇到单个数据包丢失就减半的特点，造成速率剧烈抖动，不适合传输多媒体数据。UDP由于没有拥塞机制也是不合适的。RAAR在接收端对GAIMD进行了改进，使得它有良好的速率平滑性以及能够与竞争的TCP流公平地分享带宽。RAAR算法比较简单，我们的仿真显示：RAAR的性能明显优于TFRC。由于RAAR没有每包确认机制，而且又是在接收端实现，所以它适合于升级到组播传输多媒体业务。     

Using traffic asymmetry to enhance TCP performance

A wealth of recent work has gone into optimizing the performance of Transmission Control Protocol (TCP) on the downlink channel of wireless networks such as for example, honing its congestion awareness mechanism so that it is minimally affected by random wireless losses, and optimizing achieved fairness of the end-to-end TCP rates. Other work has gone into balancing the allocation of a shared resource between the downlink and uplink in order to optimize TCP performance. We build on such previous research by proposing a cross-layer algorithm for resource allocation in OFDMA systems aiming not only to achieve optimal throughput for competing TCP flows but also to allocate resources appropriately between the downlink and uplink. This is important due to the increasing number of Internet applications where the mobile terminal is the TCP sender (social networking, peer-to-peer, etc.). Therefore, our scheme makes use of the asymmetry in the traffic and by defining the boundary between downlink and uplink capacity dynamically, enhance the TCP performance. Through numerical investigations we show the performance of the proposed scheme in terms of achieved fairness to the receivers and efficient allocation of downlink to uplink ratios based on the TCP traffic.