TCP Transfers Research Articles

A generic controller for managing TCP transfers in IEEE 802.11 infrastructure WLANs

In this paper, we present a generic controller that ensures fair and efficient operation of IEEE 802.11 infrastructure wireless local area networks (WLANs) with multiple co-channel access points. Our controller addresses performance issues of long-lived TCP transfers in multi-AP WLANs, by overlaying a coarse time-slicing scheduler on top of a cascaded fair queuing scheduler. The time slices and queue weights, used in our controller, are obtained from the solution of a constrained utility optimization formulation. A study of the impact of coarse time-slicing on TCP is also presented in this paper. We also present a methodology to improve the performance of co-existing short-lived and interactive TCP flows. Finally, we report the results of experiments performed on a real testbed, demonstrating the efficacy of our controller.

On tandem blocking queues with a common retrial queue

We consider systems of tandem blocking queues having a common retrial queue. The model represents dynamics of short TCP transfers in the Internet. Analytical results are available only for a specific example with two queues in tandem. We propose approximation procedures involving simple analytic expressions, based on mean value analysis (MVA) and on fixed point approach (FPA). The mean sojourn time of a job in the system and the mean number of visits to the orbit queue are estimated by the MVA which needs as an input the fractions of blocked jobs in the primary queues. The fractions of blocked jobs are estimated by FPA. Using a benchmark example of the system with two primary queues, we conclude that the approximation works well in the light traffic regime. We note that our approach becomes exact if the blocking probabilities are fixed. Finally, we consider two optimization problems regarding minimizing mean total sojourn time of a job in the system: (i) finding the best order of queues and (ii) allocating a given capacity among the primary queues.

Interactive wifi connectivity for moving vehicles

We ask if the ubiquity of WiFi can be leveraged to provide cheap connectivity from moving vehicles for common applications such as Web browsing and VoIP. Driven by this question, we conduct a study of connection quality available to vehicular WiFi clients based on measurements from testbeds in two different cities. We find that current WiFi handoff methods, in which clients communicate with one basestation at a time, lead to frequent disruptions in connectivity. We also find that clients can overcome many disruptions by communicating with multiple basestations simultaneously. These findings lead us to develop ViFi, a protocol that opportunistically exploits basestation diversity to minimize disruptions and support interactive applications for mobile clients. ViFi uses a decentralized and lightweight probabilistic algorithm for coordination between participating basestations. Our evaluation using a two-month long deployment and trace-driven simulations shows that its link-layer performance comes close to an ideal diversity-based protocol. Using two applications, VoIP and short TCP transfers, we show that the link layer performance improvement translates to better application performance. In our deployment, ViFi doubles the number of successful short TCP transfers and doubles the length of disruption-free VoIP sessions compared to an existing WiFi-style handoff protocol.

On the predictability of large transfer TCP throughput

Predicting the throughput of large TCP transfers is important for a broad class of applications. This paper focuses on the design, empirical evaluation, and analysis of TCP throughput predictors. We first classify TCP throughput prediction techniques into two categories: Formula-Based (FB) and History-Based (HB). Within each class, we develop representative prediction algorithms, which we then evaluate empirically over the Resilient Overlay Network (RON) testbed. FB prediction relies on mathematical models that express the TCP throughput as a function of the characteristics of the underlying network path. It does not rely on previous TCP transfers in the given path, and it can be performed with non-intrusive network measurements. We show, however, that the FB method is accurate only if the TCP transfer is window-limited to the point that it does not saturate the underlying path, and explain the main causes of the prediction errors. HB techniques predict the throughput of TCP flows from a time series of previous TCP throughput measurements on the same path, when such a history is available. We show that even simple HB predictors, such as Moving Average and Holt-Winters, using a history of few and sporadic samples, can be quite accurate. On the negative side, the accuracy of HB predictors is highly path-dependent.

TCP self-clocking and bandwidth sharing

We propose a simple queueing model for TCP transfers sharing a bottleneck link and examine its behavior when the buffer at the bottleneck is large compared to the bandwidth-delay product. This model explains some behaviors of TCP that have already been observed, and predicts other behaviors that are new. We present measurements that demonstrate these behaviors in the current Internet.

Prediction of short-lived TCP transfer latency on bandwidth asymmetric links

TCP over bandwidth asymmetric networks such as Cable TV and Asymmetric Digital Subscriber Loop (ADSL) exhibits different characteristics from TCP on symmetric links. A number of techniques have been proposed to address this problem. However, previous research has been largely focused on bulk transfers. This paper investigates the effects of bandwidth asymmetry on Web-like short-lived transfers. Two prediction models, one given in a closed form and the other in a recursive form, are presented for TCP transfers without and with ACK Filtering (AF), a representative optimizing technique for TCP transfers over bandwidth asymmetric links. Ns-2 based experiments show that these models can predict TCP transfer latency with a high degree of accuracy.

On the predictability of large transfer TCP throughput

Predicting the throughput of large TCP transfers is important for a broad class of applications. This paper focuses on the design, empirical evaluation, and analysis of TCP throughput predictors. We first classify TCP throughput prediction techniques into two categories: Formula-Based (FB) and History-Based (HB). Within each class, we develop representative prediction algorithms, which we then evaluate empirically over the RON testbed. FB prediction relies on mathematical models that express the TCP throughput as a function of the characteristics of the underlying network path. It does not rely on previous TCP transfers in the given path, and it can be performed with non-intrusive network measurements. We show, however, that the FB method is accurate only if the TCP transfer is window-limited to the point that it does not saturate the underlying path, and explain the main causes of the prediction errors. HB techniques predict the throughput of TCP flows from a time series of previous TCP throughput measurements on the same path, when such a history is available. We show that even simple HB predictors, such as Moving Average and Holt-Winters, using a history of few and sporadic samples, can be quite accurate. On the negative side, HB predictors are highly path-dependent. We explain the cause of such path dependencies based on two key factors: the load on the path and the degree of statistical multiplexing.

Analysis of link-level hybrid FEC/ARQ-SR for wireless links and long-lived TCP traffic

Since the TCP protocol uses the loss of packets as an indication of network congestion, its performance degrades over wireless links, which are characterized by a high bit error rate. Different solutions have been proposed to improve the performance of TCP over wireless links, the most promising one being the use of a hybrid model at the link level combining Forward Error Correction (FEC), Automatic Repeat Request with Selective Repeat (ARQ-SR), and an in-order delivery of packets to IP. The drawback of FEC is that it consumes some extra bandwidth to transmit the redundant information. ARQ-SR consumes extra bandwidth only when packets are lost, its drawback is that it increases the round-trip time (RTT), which may deteriorate the performance of TCP. Another drawback of ARQ-SR is that a complete packet can be retransmitted to correct a small piece of errored data. We study in this paper the performance of TCP over a wireless link implementing hybrid FEC/ARQ-SQ. The study is done by simulating and modeling long-lived TCP transfers over wireless links showing Bernoulli errors. We are motivated by how to tune link-level error recovery, e.g. amount of FEC, persistency of ARQ, so as to maximize the performance of TCP. We provide results for different physical characteristics of the wireless link (delay, error rate) and for different traffic loads (number of TCP connections).

Estimating loss rates with TCP

Estimating loss rates along a network path is a problem that has received much attention within the research community. However, deriving accurate estimates of the loss rate from TCP transfers has been largely unaddressed. In this paper, we first show that using a simple count of the number of retransmissions yields inaccurate estimates of the loss rate in many cases. The mis-estimation stems from flaws in TCP's retransmission schemes that cause the protocol to spuriously retransmit data in a number of cases. Next, we develop techniques for refining the retransmission count to produce a better loss rate estimate for both Reno and SACK variants of TCP. Finally, we explore two SACK-based variants of TCP with an eye towards reducing spurious retransmits, the root cause of the mis-estimation of the loss rate. An additional benefit of reducing the number of needless retransmits is a reduction in the amount of shared network resources used to accomplish no useful work.

A performance comparison of Mobile IPv6, Hierarchical Mobile IPv6, fast handovers for Mobile IPv6 and their combination

Mobile IP, the current IETF proposal for IP mobility support, represents a key element for future All-IP wireless networks to provide service continuity while on the move within a multi-access environment. We conducted a performance evaluation of Mobile IPv6 and its proposed enhancements, i.e., Fast Handovers for Mobile IPv6, Hierarchical Mobile IPv6 and our proposed combination of them, using the network simulator ns-2 for the case of a 'hot spot' deployment scenario. The simulation scenario comprises four access routers and up to 50 mobile nodes that communicate in accordance with the IEEE 802.11 wireless LAN standard. The study provides quantitative results of the performance improvements obtained by the proposed enhancements as observed by a single mobile user with respect to handoff latency, packet loss rate and achieved bandwidth per station. As a complementary part of the study, the signaling load costs associated with the performance improvements provided by the enhancements has been analyzed. The simulation environment allowed us also to investigate the behavior of the protocol in extreme cases, e.g., under channel saturation conditions and considering different traffic sources: CBR, VoIP, Video and TCP transfers. While some simulation results corroborate the intention of the protocols specifications, other results give insights not easily gained without performing simulations. This study provides a deep understanding of the overall performance of the various protocols and supports the design process of a Mobile IPv6-based network when a decision of whether it is appropriate to implement any of the proposed Mobile IPv6 enhancements has to be made.

Socket Buffer Auto-Sizing for High-Performance Data Transfers

It is often claimed that TCP is not a suitable transport protocol for data intensive Grid applications in high-performance networks. We argue that this is not necessarily the case. Without changing the TCP protocol, congestion control, or implementation, we show that an appropriately tuned TCP bulk transfer can saturate the available bandwidth of a network path. The proposed technique, called SOBAS, is based on automatic socket buffer sizing at the application layer. In non-congested paths, SOBAS limits the socket buffer size based on direct measurements of the received throughput and of the corresponding round-trip time. The key idea is that the send window should be limited, after the transfer has saturated the available bandwidth in the path, so that the transfer does not cause buffer overflows (‘self-induced losses’). A difference with other socket buffer sizing schemes is that SOBAS does not require prior knowledge of the path characteristics, and it can be performed while the transfer is in progress. Experimental results in several high bandwidth-delay product paths show that SOBAS provides consistently a significant throughput increase (20% to 80%) compared to TCP transfers that use the maximum possible socket buffer size. We expect that SOBAS will be mostly useful for applications such as GridFTP in non-congested wide-area networks.

An integrated model for the latency and steady-state throughput of TCP connections

Most TCP connections in today’s Internet transfer data on the order of only a few kilobytes. Such TCP transfers are very short and spend most of their time in the slow start phase. Thus the underlying assumptions made by steady-state models cease to hold, making them unsuitable for modeling finite flows. In this paper, we propose an accurate model for estimating the transfer times of TCP flows of arbitrary size. Our model gives a more accurate estimation of the transfer times than those predicted by Cardwell et al. [Proceedings of the IEEE INFOCOM, Tel Aviv, Israel, March 2000, pp. 1742–1751], which extends the steady-state analysis of Padhye et al. [IEEE/ACM Trans. Networking 8 (2) (2000) 133] to model finite flows. The main features of our work are the modeling of timeouts and slow start phases which occur anywhere during the transfer and a more accurate model for the evolution of the cwnd in the slow start phase. Additionally, the proposed model can also model the steady-state throughput of TCP connections. The model is verified using web based measurements of real life TCP connections. We also introduce an empirical model which allows a better “feel” of TCP latency and the nature of its dependence on loss probabilities and window limitation. Finally, the paper investigates the effect on window limitation and packet size on TCP latency.

An evaluation of TCP with larger initial windows

TCP's slow start algorithm gradually increases the amount of data a sender injects into the network, which prevents the sender from overwhelming the network with an inappropriately large burst of traffic. However, the slow start algorithm can make poor use of the available bandwidth for transfers which are small compared to the bandwidth-delay product of the link, such as file transfers up to few thousand characters over satellite links or even transfers of several hundred bytes over local area networks. This paper evaluates a proposed performance enhancement that raises the initial window used by TCP from 1 MSS-sized segment to roughly 4 KB. The paper evaluates the impact of using larger initial windows on TCP transfers over both the shared Internet and dialup modem links.

Improving the start-up behavior of a congestion control scheme for TCP

Based on experiments conducted in a network simulator and over real networks, this paper proposes changes to the congestion control scheme in current TCP implementations to improve its behavior during the start-up period of a TCP connection.The scheme, which includes Slow-start, Fast Retransmit, and Fast Recovery algorithms, uses acknowledgments from a receiver to dynamically calculate reasonable operating values for a sender's TCP parameters governing when and how much a sender can pump into the network. During the start-up period, because a TCP sender starts with default parameters, it often ends up sending too many packets and too fast, leading to multiple losses of packets from the same window. This paper shows that recovery from losses during this start-up period is often unnecessarily time-consuming.In particular, using the current Fast Retransmit algorithm, when multiple packets in the same window are lost, only one of the packet losses may be recovered by each Fast Retransmit; the rest are often recovered by Slow-start after a usually lengthy retransmission timeout. Thus, this paper proposes changes to the Fast Retransmit algorithm so that it can quickly recover from multiple packet losses without waiting unnecessarily for the timeout. These changes, tested in the simulator and on the real networks, show significant performance improvements, especially for short TCP transfers. The paper also proposes other changes to help minimize the number of packets lost during the start-up period.