Mahdavi, J., Floyd, S.: TCP-friendly unicast rate-based flow control (1997) Note sent to the end2end-interest mailing list.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....state throughput of a ow along a path to the throughput that TCP would achieve along that path. For TCP trac, a great deal of work has been done to determine the equation expressing the throughput as a function of the packet size, the packet loss rate, and the round trip time along that path [11], 8] 9] 14] It has been advocated that any new congestion control algorithm should exhibit the same steady state ow rate as suggested by the TCP equation [11] 2] ensuring that the ows will then share available bandwidth fairly across a bottleneck link, since across the bottleneck link ....

.... equation expressing the throughput as a function of the packet size, the packet loss rate, and the round trip time along that path [11] 8] 9] 14] It has been advocated that any new congestion control algorithm should exhibit the same steady state ow rate as suggested by the TCP equation [11], 2] ensuring that the ows will then share available bandwidth fairly across a bottleneck link, since across the bottleneck link We review the root causes of large IGMP leave latency in Section III A. both streams experience the same packet loss rate. In addition, there is an emerging body ....

[Article contains additional citation context not shown here]

J. Mahdavi and S. Floyd. TCP-friendly Unicast Rate-based Flow Control. Note sent to end2end-interest mailing list, January 1997.

....rate control is TFRC [10] which we use most of the time in this paper as a recurring example. Because is assumed to represent TCP s loss throughput equation, it is expected that such a rate control is TCP friendly, i.e. our adaptive source shares the network fairly with competing TCP sources [11]. More precisely, this is required to happen on two time scales: in the short run, response to congestion should be commensurate to that of TCP; in the long run, average throughput should not exceed that of TCP. In this paper, we focus on the latter (we call it long run TCP friendly ) we refer ....

....More precisely, this is required to happen on two time scales: in the short run, response to congestion should be commensurate to that of TCP; in the long run, average throughput should not exceed that of TCP. In this paper, we focus on the latter (we call it long run TCP friendly ) we refer to [11, 7, 8] for some definitions regarding the concept of TCP friendliness. Our goal is to identify the key factors that drive whether, and how far, the equation based rate control is long run TCP friendly. We first point out that TCP and our source may experience different loss event ratios, and thus ....

Jamshid Mahdavi and Sally Floyd. TCP-Friendly Unicast Rate-based Flow Control. Technical note sent to end-2-end interest mailing list, http://www.psc.edu/networking/papers/tcp friendly.html, January 1997.

....that saves on the processing of packet events for TCP flows. Needless to say, TCP traffic represents a large portion of the Internet traffic today and, therefore, large scale network simulations typically involve the simulation of a large number of TCP packet flows. It is well established[4, 5] that the average behavior of a TCP flow is predictable given a steady state path condition. It is our goal in this paper to exploit this fact to predict the average behavior of a TCP flow over a future period of time where steady state conditions hold, thus allowing for a reduction (or ....

....as a function of loss ratio, round trip time, b (number of segments ACKed by one ACK) and the maximum window size advertised by the receiver. The B(loss, RTT , b, Wmax ) formula applies to a bulk transfer TCP flow, i.e. a flow with an unlimited amount of data to send. Compared to previous work in [4], this model captures the timeout events in the steady state and hence the applicability of this equation is not limited to small packet loss ratio. With formula based prediction, we measure for each flow its average RTT and the packet loss probability on its path. We then estimate the average ....

J.Mahdavi and S. Floyd. TCP-Friendly Unicast Rate-Based Flow Control. Sent to end2end-interest mailing list, Jan 1997.

....of the measured loss rate. Bansal et al. 4] investigate the behavior of slowly responsive congestion control algorithms. However, to be safe for deployment in the Internet these algorithms must be stable and interact well with TCP. A recently proposed solution is the TCP compatible paradigm [18]. A congestion control mechanism is TCPcompatible if its bandwidth usage, in the presence of a constant loss rate, is the same as that of TCP. Our work. We suggest an adaptive AIMD scheme, in which the parameters may change dynamically with respect to the obtained feedback . In a nutshell, ....

J. Mahdavi and S. Floyd, \TCP-Friendly Unicast Rate-Based Flow Control," Available from http://www.psc.edu/ networking/papers/tcp friendly.html, Jan. 1997.

....window has the same size as the amount of bu#er space (B) available in the receiver, yielding an average throughput (T ) of: 4.1) where RTT is the dynamic estimate of the round trip time. When the window size is changing due to the congestion avoidance algorithm [61, 121] Mahdavi and Floyd [79], Mathis et.al. 81] have derived and verified expressions relating TCP throughput to the packet loss rate (p) T = k MSS (4.2) where MSS, the maximum segment size, is the amount that the TCP sending window 63 increases per RTT when there is no packet loss. k is a constant between 0.7 [81] ....

....et.al. 81] have derived and verified expressions relating TCP throughput to the packet loss rate (p) T = k MSS (4.2) where MSS, the maximum segment size, is the amount that the TCP sending window 63 increases per RTT when there is no packet loss. k is a constant between 0.7 [81] and 1. 3 [79] depending on the particular derivation of Equation 4.2. A non TCP flow sharing the same path as a TCP flow will experience and measure similar network conditions. By choosing B and MSS, and measuring p and RTT appropriately, it is possible for the non TCP flow to estimate the throughput of the ....

J. Mahdavi and S. Floyd. Tcp-friendly unicast rate-based flow control. Technical note sent to the end2end-interest mailing list, 1997.

....TCP flows. In this case, the video transmission may take an unfair proportion of the bandwidth. Thus, there is a need to match the video data rate to the sending rate that a TCP flow would send if it was in the same situation as the video flow, that is, the video sending rate must be TCP friendly [29]. However, TCP dictates a sending rate that is rapidly oscillating according to congestion, whereas VBR compressed video bit rate oscillates according to compression efficiency. One common approach has been to send data at TCP s average sending rate [2] 3] Since the averaging leads to slow ....

J. Mahdavi and S. Floyd, "TCP-friendly unicast rate-based flow control," Note sent to the end2end-interest mailing list, 1997.

....with each other and with current TCP based applications. One way to ensure this is to implement some form of congestion control that adapts the transmission rate in a way that fairly shares congested bandwidth with TCP applications, thus falling into the category of TCP friendly congestion control [25]. There has been significant previous research on TCPFriendly control mechanisms and many control schemes were proposed in the literature. Prominent examples of these schemes are: 1) window based control mechanisms such as TEAR [26] 2) mechanisms based on additive increase, multiplicative ....

J. Mahdavi and S. Floyd, "TCP-Friendly unicast rate-based flow control," in Draft posted on end2end mailing list, Jannuary 1997, http://www.psc.edu/networking/papers/tcp friendly.html.

....end toend congestion control that results in bursty network traffic coupled with the delivery of unstable QoS over time. Researchers have devised multiple solutions to the substantial delays incurred by the transfer of large data across wide area networks. Recent work on TCP friendly communication[9] and on dynamic buffer right sizing[4] attempts to reduce bursty traffic behavior by controlling the ways in which applications or protocols send data across the network. Specifically, while the AIMD (Additive Increase Multiplicative Decrease) algorithm used by the dominating transport protocol ....

.... by the dominating transport protocol TCP is shown to converge to a fair state in terms of bandwidth usage, its adaptation behavior makes TCP traffic bursty in nature and delivers unstable QoS over time, which is particularly problematic for large data transfer[4, 15] and for real time applications[9]. TCP friendly congestion control algorithms[9, 14] focus on achieving a relatively stable sending rate that is adapted to the network s level of congestion. Improved bandwidth can also be realized by concurrent use of multiple routes or even sockets[15] An alternative solution is to use ....

[Article contains additional citation context not shown here]

J. Mahdavi and S. Floyd. TCP-friendly unicast ratebased flow control, Jan 1997.

.... area has shown that the stationary throughput of a saturated TCP sender (i.e. one with an infinite amount of bytes to send) is on the inverse order of the square root of the loss rate observed by the connection [17] In the following we consider the so called MF model, due to Madhavi and Floyd [18]. Another model of interest is the PFTK one, which was proposed by Padhye, Firoiu, Towsley and Kurose in [19] Mahdavi and Floyd propose in [18] the equation : BW = 1:22 Theta MTU RTT Theta Loss (10) to compute the bandwidth that a TCP connection receives, provided it has MTU bytes per ....

.... order of the square root of the loss rate observed by the connection [17] In the following we consider the so called MF model, due to Madhavi and Floyd [18] Another model of interest is the PFTK one, which was proposed by Padhye, Firoiu, Towsley and Kurose in [19] Mahdavi and Floyd propose in [18] the equation : BW = 1:22 Theta MTU RTT Theta Loss (10) to compute the bandwidth that a TCP connection receives, provided it has MTU bytes per packet, and incurs a roundtrip time of RTT seconds and a loss rate of Loss. This equation comes from a steady state analysis of the TCP congestion ....

[Article contains additional citation context not shown here]

J. Mahdavi and S. Floyd. TCP-friendly unicast rate-based flow control. Technical note sent to the end2endinterest mailing list, january 8 1997.

....with each other and with current TCP based applications. One way to ensure this is to implement some form of congestion control that adapts the transmission rate in a way that fairly shares congested bandwidth with TCP applications, thus falling into the category of TCP friendly congestion control [25]. There has been significant previous research on TCPFriendly control mechanisms and many control schemes were proposed in the literature. Prominent examples of these schemes are: 1) window based control mechanisms such as TEAR [26] 2) mechanisms based on additive increase, multiplicative ....

J. Mahdavi and S. Floyd, "TCP-Friendly unicast rate-based flow control," in Draft posted on end2end mailing list, Jannuary 1997, http://www.psc.edu/networking/papers/tcp friendly.html.

....permits the video to be VBR encoded, although the results remain insightful for the special case of CBR video. The model supposes that the bandwidth available to the video streaming application is variable; it could, for example, be the fair share bandwidth determined by a TCP friendly algorithm [7 9]. We also suppose that the video is layered encoded. Layered encoding is useful in order to cope with the heterogeneity of user access rates and with the competing traffic in the links between server and client. In this paper we suppose that the video is encoded in two layers a base layer and ....

....designed to be cooperative with the TCP connections by reacting to congestion [20] This can be done, for example, by probing to discover the fair share of network bandwidth and transmit at a rate that does not exceed this fair share. Applications with this property are said to be TCP friendly [7, 8]. An application s fair share rate can be estimated by its round trip times and its loss rates [7 9] A. Related Research Rejaie et al. [21, 22] consider a broad range of architectural issues for streaming layered encoded video. They argue for the need for end to end congestion control, quality ....

[Article contains additional citation context not shown here]

J. Mahdavi and S. Floyd, "TCP-Friendly Unicast Rate-Based Flow Control, " Tech. Rep., Jan. 1997.

....applications become widespread on the Internet, the problem of streaming stored multimedia over the best effort Internet while remaining fair to competing TCP traffic has become increasingly important. The need to be fair to competing TCP traffic has lead to the notion of TCP friendly streams [4, 2, 6, 8]. The available bandwidth to a TCP friendly scheme varies significantly over both medium and long time scales [10] For streaming stored video, it is therefore useful to consider schemes that adapt video quality as a function of the available bandwidth. At the same time, the scheme should not lead ....

J. Mahdavi and S. Floyd. TCP-Friendly Unicast Rate-Based Flow Control. Technical report, Jan. 1997.

.... of customers on the route of a CAC (call acceptance controller) probe stream can also be estimated for a product form Kelly network [7] The steady state throughput of a bulk transfer TCP flow can be estimated as a function of loss rate and round trip time using the so called TCP friendly formula [8], 9] 10] Although the previous estimates have been devised under the assumption that there exists a single bottleneck link on a path connecting two hosts, experiments reported in the previous references indicate that these estimates still perform reasonably well when this assumption is ....

J. Mahdavi and S. Floyd, "TCP-friendly unicast rate-based flow control," January 1997, Technical note sent to the end2end-interest mailing list, available via http://www.psc.edu/networking/papers/tcp_friendly.html.

....studies to understand the behavior of Internet applications. In particular, special attention has been paid to modeling TCP friendly protocols in the presence of packet loss, and especially to deriving the throughput of a TCP bulk data transfer. In the absence of timeouts, it was observed in [4, 3, 7] that the throughput of a TCP connection is given by = c ; 1) where c is an empirical constant taking values 1.22 or 1.31, depending on whether the receiver uses delayed acknowledgements or not, R is the round trip time experienced by the packets the connection, and p is the the expected ....

....constant c = 1:31 appears in this paper as a universal constant, which is a characteristics of the additive increase multiplicative decrease scheme performed in congestion avoidance. Note that 1:31 is really the constant appearing in equation (1) according to the experiments and simulations (see [4] for example) and not the constant 3=2 as suggested by equation (2) The organization of this paper is as follows: In Section 2, the model is described and upper and lower bounds for the throughput of a TCP connection experiencing an arbitrary constant loss probability are obtained. In ....

[Article contains additional citation context not shown here]

J. Madhavi and S. Floyd. TCP friendly unicast rate-based flow control. end2end-interest mailing list. January 1997.

....Generally, we only send feedback if t TM . If NREQ = 0, we send feedback if t T I or jR(t) R(t0 )j R(t) #. If NREQ 0, we send feedback if t TR or NREQ N thres . 2.2. 5 Congestion control In order to avoid interfering with congestion controlled trac (e.g. TCP) in an unfair way [28], and to prevent starvation among competing reservations, SRP senders must respond to congestion by limiting the rate at which they send request packets. A mechanism similar to TCP congestion control can also be used with SRP: the sender maintains a congestion window, which grows while request ....

Floyd, Sally; Mahdavi, Jamshid. TCP-Friendly Unicast Rate-Based Flow Control, http://www.psc.edu/networking/papers/tcp_friendly.html, Technical note, January 1997.

.... FEC is too often used without rate control, what leads to more congestion, loss and then worse audio quality [6] The purpose of this work is to add adaptive FEC to an existing software: the Robust Audio Tool [7] FEC will be constrained by a TCP friendly rate, proposed by Mahdavi and Floyd [16]. In this paper the general problem of optimizing quality at reception is presented, a more speci c solution is deduced. 2 State of the Art 2.1 FEC De nition Forward Error Correction (FEC) relies on the addition of repair data to a stream, from which the content of lost packets may be ....

.... can be added when loss occurs, but in the same time the source encoding must be changed to use less bandwidth in response to congestion [6] Our framework is to continuously ensure that the overall rate transmission is smaller or equal than a TCP friendly rate proposed by Mahdavi and Floyd [16]. It is a combined source rate redundancy control. 2.2 TCP Friendly Rate Control As networked multimedia applications become widespread, it becomes increasingly important to ensure that they can share resources fairly with each other and with current TCP based applications, the dominant source ....

[Article contains additional citation context not shown here]

J. Mahdavi and S. Floyd. TCP-Friendly Unicast Rate-Based Flow Control. Draft posted on end2end mailing list, Jannuary 1997. http://www.psc.edu/papers/tcp friendly.html.

....feedback. In reality however, the probability to receive a congestion indication is higher for the user with a larger load. Subsequent analytical studies of TCP congestion control represent feedback more realistically and predict the transmission rate for a TCP connection more accurately [2] [10], 12] 13] Furthermore, experiments and more realistic models with proportional negative feedback agree that bandwidth allocation under TCP does not converge to maxmin fairness [13] 16] While reliance of TCP on AIMD does not attain the original goal of convergence to maxmin fairness, it is ....

J. Mahdavi and S. Floyd. TCP-Friendly Unicast RateBased Flow Control. End2end-interest mailing list, January 1997.

.... FEC is too often used without rate control, what leads to more congestion, loss and then worse audio quality [6] The purpose of this work is to add adaptive FEC to an existing software: the Robust Audio Tool [7] FEC will be constrained by a TCP friendly rate, proposed by Mahdavi and Floyd [16]. The general problem of optimizing quality at reception is presented, a more specific solution is deduced. A short presentation of software s architecture should help people interested in improving this work. 2 State of the art This section is a survey of useful techniques for audio exchange ....

.... can be added when loss occurs, but in the same time the source encoding must be changed to use less bandwidth, in response to congestion [6] Our framework is to continuously ensure that the overall rate transmission is smaller or equal than a TCP friendly rate proposed by Mahdavi and Floyd [16]. It is a combined source rate redundancy control. Two important parameters concerning SFEC are: k denotes the number of copies for the same audio unit. In the figure 1, for example, k = 2. k is bound to the loss probability of the network and the the desired apparent loss rate after ....

[Article contains additional citation context not shown here]

J. Mahdavi and S. Floyd. TCP-Friendly Unicast Rate-Based Flow Control. Draft posted on end2end mailing list, Jannuary 1997. http://www.psc.edu/papers/tcp friendly.html. 13

....to improving TCP throughput are based on standard ideas that do not impact assumptions about fairness when competing with other traffic in the network [20, 43] Application based approaches are often more aggressive about acquiring network capacity. Examples include most non TCP based transports [14, 32, 36] and all parallel TCP approaches [19, 25, 44, 49] This paper describes a daemon based mechanism to implement both standard and non standard TCP adjustments for individual flows using network metrics across designated paths. No modifications to the application are required, and the user does not ....

J. Mahdavi and S. Floyd. TCP-Friendly Unicast Rate-Based Flow Control, 1997. URL: http://www.psc.edu/networking/papers/tcp friendly.html.

....using TCP s absolute linear increase takes much too long and results in lowered network utilization. In this particular case, convergence can take as long as (100 Mb 50 Mb) 1500 B RTT. 8 b B) 4,168 RTTs or (4,168 RTTs. 100 ms RTT) 416.8 seconds = 6. 947 minutes In 1997, Mahdavi and Floyd [13] informally proposed the notion of equationbased congestion control for unicast applications. While the AIMD algorithm found in TCP backs off by cutting its sending rate in half in response to a single congestion indication, equation based congestion control uses a control equation that more ....

J. Mahdavi and S. Floyd. TCP-friendly unicast rate-based flow control. Technical report, Note sent to end2end-interest mailing list, 1997.

....with each other and with current TCP based applications. One way to ensure this is to implement some form of congestion control that adapts the transmission rate in a way that fairly shares congested bandwidth with TCP applications, thus falling into the category of TCP friendly congestion control [24]. There has been significant previous research on TCP Friendly control mechanisms and many control schemes were proposed in the literature. Prominent examples of these schemes are: 1) window based control mechanisms such as TEAR [25] 2) mechanisms based on additive increase, multiplicative ....

J. Mahdavi and S. Floyd, "TCP-Friendly unicast rate-based flow control," in Draft posted on end2end mailing list, Jannuary 1997, http://www.psc.edu/networking/papers/tcp_friendly. html.

....the flow control problem in decision theoretic terms, however, as we discuss in Section II, setting a good optimization criteria may be tricky. Related work: Yang and Lam [4] studied AIMD flow control and derived the relationship between the parameters so that the flow would be TCP friendly [5]. Bansal and Balakrishnan [6] evaluate a class of non linear congestion control algorithms. Both suggest different update parameters for the flow control, however, the parameters do not change over time. In contrast, we suggest that the algorithm modify the parameters dynamically. The rest of this ....

M. Floyd, "Tcp-friendly unicast rate-based flow control," 1997.

....Sisalem and Schulzrinne s Loss Delay Adjustment (LDA) algorithm [3] uses an additive increase multiplicative decrease (AIMD) rate control based on RTCP feedbacks. Rate Adaptation Protocol (RAP) by Rejaie et al. [4] adjusts its rate in AIMD fashion based on frequent receiver acknowledgements. In [5,6,7], algorithms based on derived throughput equations from more detailed TCP models are proposed. In common, these rate based approaches measure the model parameters such as packet loss rate and round trip time periodically and adjust the rate based on throughput equations or other strategies. A ....

J. Mahdavi and S. Floyd. TCP-friendly unicast rate based flow control, January1997. http://www.psc.edu/network-ing/papers/tcp friendly.html.

.... delay, then 0 where 1:31: In other words, we recover the asymptotics : Surprisingly, while the value 1:31 for the constant is derived solely from the structure of the problem, it is nevertheless in agreement with the values obtained from experiments and simulations [33]. On the other hand, ad hoc approximations suggest the values 3=2 1:22 [3] or 2 [7] for this constant. This evidence provides one more data point in support of the robustness of traffic modeling via limit theorems. In a different vein, Sharma and Purkayastha [34] consider an ODE ....

J. Mahdavi and S. Floyd, "TCP-friendly unicast rate-based flow control," Technical Note sent to the end2end-interest mailing list, Jan. 1997, Available from http://www.psc.edu/networking/papers/tcp friendly.html.

....in TCP congestion control. According to the stochastic analysis by Padhye et al., the throughput of a TCP connection at a steady state is inversely proportional to RTT [10] Furthermore, TCP friendliness has emerged as a new notion of fairness where users with longer RTTs have smaller fair loads [9, 11]. Our truly intriguing result is that AIMD has multiple attractors and can converge to steady states where users with longer RTTs impose larger loads. This contradicts Padhye formula for the steady state TCP throughput. How can we resolve the contradiction Attributing it to the assumption of ....

J. Mahdavi and S. Floyd. TCP-Friendly Unicast Rate-Based Flow Control. End2end-interest mailing list, January 1997.

....(LDA) scheme [20] that uses an AIMD rate control at the sender and deploys RTP [18] for feedback. However, to be safe for deployment in the Internet the alternative congestion control protocols must be stable and interact well with TCP. A recently proposed solution is the TCP compatible paradigm [11]. The cornerstone of this approach is the observation, that in the presence of a constant packet loss rate p, one can characterize the bandwidth usage of a TCP ow as proportional to 1= p p. A congestion control mechanism is TCP compatible if its bandwidth usage, in the presence of a constant ....

....state the bandwidth utilization is close to the optimal; 2) the time it takes for a connection to reach its fair share is proportional to the square of the logarithm of the link bandwidth. This scheme, as we will show, is not compatible with canonical TCP that is de ned by Madhavi and Floyd [11]. When competing with TCP for bottleneck resources, it will consume larger fraction of bandwidth. However, following Bansal and Balakrishnan [2] we believe that the primary threat to the Internet stability comes not from the ows using some form of TCP incompatible congestion control but from ....

J. Mahdavi and S. Floyd, \TCP-Friendly Unicast Rate-Based Flow Control," Available from http://www.psc.edu/ networking/papers/tcp friendly.html, Jan. 1997.

.... many real time applications would find halving the sending rate of a flow to be too severe a response to a congestion indication, as it can noticeably reduce the flow s user perceived quality [25] In the past few years, many unicast congestion control schemes have been proposed and investigated [12, 16, 28, 23, 3, 18, 22, 25, 20, 29, 9]. The common objective of these studies is to find a good alternative to the congestion control scheme of TCP. Since the dominant Internet traffic is TCP based [26] it is important that new congestion control schemes be TCP friendly. By this, we mean that the sending rate of a non TCP flow should ....

....Since the dominant Internet traffic is TCP based [26] it is important that new congestion control schemes be TCP friendly. By this, we mean that the sending rate of a non TCP flow should be approximately the same as that of a TCP flow under the same conditions of round trip time and packet loss [16]. The congestion control schemes investigated can be divided into two categories: AIMD based [12, 23, 3, 22, 18] and formula based [16, 28, 25, 20, 29, 9] Roughly speaking, AIMD based schemes emulate the increase by one and decrease to half window behavior of TCP. Formula based schemes use a ....

[Article contains additional citation context not shown here]

J. Mahdavi and S. Floyd. TCP-friendly unicast rate-based flow control. Technical note sent to the end2end-interest mailing list, 1997.

....the dominant Internet traffic is TCP based [23] it is important that new congestion control schemes be TCP friendly. By this, we mean that the sending rate of a non TCP flow should be approximately the same as that of a TCP flow under the same conditions of round trip time and packet loss rate [16, 4]. However, evaluations of these new protocols have been focused mainly on fairness. Two methods were proposed to establish the fairness of a protocol. The first is the Chiu and Jain s phase space method [7] which can be used to show that a protocol will converge asymptotically to a fair state, ....

....paper is to evaluate these properties by analytically and experimentally studying the transient behaviors of several TCP friendly congestion control protocols. Proposed congestion control schemes in the literature fall into two major categories: AIMD based [13, 21, 6, 19, 27, 20] and formula based [16, 24, 22, 17, 10]. For our study, we select TCP [14] and GAIMD [27] as representatives of the AIMD based schemes. GAIMD generalizes TCP by parameterizing the congestion window increase value and decrease ratio. That is, in the congestion avoidance state, the window size is increased by per window of packets ....

J. Mahdavi and S. Floyd. TCP-friendly unicast rate-based flow control. Note sent to the end2end-interest mailing list, 1997.

....fairly with each other and with current TCPbased applications. One way to ensure this is to implement some form of congestion control that adapts the transmission rate in a way that fairly shares congested bandwidth with TCP applications. One definition of fair is that of TCP friendliness [24] if a non TCP connection shares a bottleneck link with TCP connections, traveling over the same network path, the non TCP connection should receive the same share of bandwidth as a TCP connection. There has been significant previous research on TCP Friendly control mechanisms and many control ....

....to be unnecessarily severe (as it can noticeably decrease the user perceived quality [30] For this reason, equation based control mechanism seem to be leading candidates for mechanisms to provide relatively smooth congestion control for real time traffic. In Equation based congestion control [24] schemes, the sender uses an equation characterizing the allowed sending rate of a TCP connection as a function of the RTT and packet loss rate, and adjusts its sending rate according to those measured parameters. A key issue, when using these schemes, is of course to choose a reliable ....

[Article contains additional citation context not shown here]

J. Mahdavi and S. Floyd, "TCP-Friendly unicast rate-based flow control," in Draft posted on end2end mailing list, Jannuary 1997, http://www.psc.edu/networking/papers/tcp friendly.html.

....the TCP sending rate in congestion avoidance state is controlled by a congestion window which is halved for every window of data containing a packet drop, and increased by one packet per window of data acknowledged. Recently, many new congestion control protocols were proposed and investigated [4, 6, 13, 11, 1, 14, 7, 10, 12, 8, 15]. The objective of these new congestion protocols is to address the needs of new multimedia applications. We notice that, like TCP, many of these proposals are also based on the AIMD principle. Further, there is even a common belief that AIMD is optimal and is a necessary condition for a ....

J. Mahdavi and S. Floyd. TCP-friendly unicast ratebased flow control. Note sent to the end2end-interest mailing list, 1997.

No context found.

Mahdavi, J., Floyd, S.: TCP-friendly unicast rate-based flow control (1997) Note sent to the end2end-interest mailing list.

No context found.

J. Mahdavi and S. Floyd, TCP-Friendly Unicast RateBased Flow Control. Technical note sent to the end2endinterest mailing list, Jan 1997. http://www.psc.edu/networking/papers/tcpfriendly.html.

No context found.

J. Mahdavi and S. Floyd. TCP-friendly unicast rate-based flow control. Note sent to the end2end-interest mailing list, Jan. 1997.

No context found.

J. Mahdavi and S. Floyd. TCP-friendly unicast rate-based flow control. Note sent to the end2end-interest mailing list, 1997.

No context found.

J. Mahdavi and S. Floyd, "TCP-Friendly Unicast Rate-based Flow Control", Technical Note to End2end-interest Mailing List, Jan. 1997.

No context found.

J. Mahdavi and S. Floyd, "TCP-friendly unicast rate-based flow control, " note sent to end2end-interest mailing list, Jan. 1997.

No context found.

J. Mahdavi and S. Floyd. TCP-Friendly Unicast RateBased Flow Control. End2end-interest mailing list, January 1997.

No context found.

J. Mahdavi and S. Floyd, "TCP-friendly unicast rate-based flow control," Technical note sent to the end2end-interest mailing list, January 1997.

No context found.

J. Mahdavi and S. Floyd. TCP-Friendly Unicast Rate-Based Flow Control. End2end-interest mailing list, January 1997.

No context found.

J. Mahdavi and S. Floyd, "Tcp-friendly unicast rate-based flow control. " Note sent to end2end-interest mailing list, Jan. 1997.

No context found.

J. Mahdavi and S. Floyd. TCP-Friendly Unicast Rate-Based Flow Control, Jan. 1997.

No context found.

J. Mahdavi and S. Floyd, "TCP-Friendly Unicast Rate-based Flow Control ", Technical Note to End2end-interest Mailing List, Jan. 1997.

No context found.

J. Mahdavi and S. Floyd. "TCP-friendly unicast rate-based flow control ". Technical report, Technical note sent to the end2end-interest mailing list, 1997.

No context found.

J. Mahdavi and S. Floyd. TCP-friendly unicast rate-based flow control. Note sent to the end2end-interest mailing list, 1997.

No context found.

J. Mahdavi and S. Floyd. TCP-friendly unicast rate-based flow control. Technical Note sent to the end2end-interest mailing list, January 1997. Available from http://www.psc.edu/ networking/papers/tcp_friendly.html.

No context found.

Jamshid Mahdavi and Sally Floyd. Tcpfriendly unicast rate-based flow control. Technical note sent to end-2-end interest mailing list, http://www.psc.edu/networking/papers/tcp friendly.html, January 1997.

No context found.

J. Mahdavi and S. Floyd. "TCP-friendly unicast rate-based flow control". Note sent to end2end-interest mailing list, Jan 1997.

No context found.

Mahdavi, and S.Floyd "TCP-Friendly Unicast Rate-Based Flow Control", Technical note sent to the end2end-interest mailing list, Jan., 1997.

No context found.

J. Mahdavi and S. Floyd. TCP-Friendly Unicast Rate-Based Flow Control. End2end-interest mailing list, January 1997.

No context found.

J. Mahdavi, and S. Floyd, "TCP-Friendly Unicast Rate-Based Flow Control," Techical Note on End2end Mailing List, January 1997.

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC