No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
On Estimating End-to-End Network Path Properties.
Abstract
The more information about current network conditions available to a transport protocol, the more efficiently it can be use the network to transfer its data. In networks such as the Internet, the transport protocol must often form its own estimates of network properties based on measurements performed by the connection endpoints. We consider two basic transport estimation problems: determination the setting of the retransmission timer (RTO) for a reliable protocol, and estimating the bandwidth available to a connection as it begins. We look at both of these problems in the context of TCP, using a large TCP measurement set [Pax97b] for trace-driven simulations. For RTO estimation, we evaluate a number of different algorithms, finding that the performance of the estimators is dominated by their minimum values, and to a lesser extent, the timer granularity, while being virtually unaffected by how often round-trip time measurements are made or the settings of the parameters in the exponentially-weighted moving average estimators commonly used. For bandwidth estimation, we explore techniques previously sketched in the literature [Hoe96, AD98] and find that in practice they perform less well than anticipated. We then develop a receiver-side algorithm that performs significantly better.
... Moreover, a conservative RTO policy setting large values will also decrease throughput because lost packets will not be retransmitted until their RTOs expire, rendering networks sluggish and inefficient [14]. ...
... In TCP/IP, the Jacobson algorithm [13] is used to estimate the RTO, and several analyses have confirmed its efficiency in this architecture [14,17]. However, in-network caching, request aggregation, and dynamic forwarding of interest packets in CCN make its performance in this architecture controversial [5,6,16,18]. ...
... The accuracy of the Jacobson algorithm depends highly upon parameter settings, and several experimental evaluations have assigned them in TCP [14,27,28]. However, the performance of Jacobson algorithm in CCN with a significant architectural difference has not yet been evaluated, and the estimator's proper setting in CCN remains unknown. ...
Accurately estimating of Retransmission TimeOut (RTO) in Content-Centric Networking (CCN) is crucial for efficient rate control in end nodes and effective interface ranking in intermediate routers. Toward to this end, the Jacobson algorithm, which is an Exponentially Weighted Moving Average (EWMA) on the Round Trip Time (RTT) of previous packets, is a promising scheme. Assigning the lower bound to RTO, determining how an EWMA rapidly adapts to changes, and setting the multiplier of variance RTT have the most impact on the accuracy of this estimator for which several evaluations have been performed to set them in Transmission Control Protocol/Internet Protocol (TCP/IP) networks. However, the performance of this estimator in CCN has not been explored yet, despite CCN having a significant architectural difference with TCP/IP networks. In this study, two new metrics for assessing the performance of RTO estimators in CCN are defined and the performance of the Jacobson algorithm in CCN is evaluated. This evaluation is performed by varying the minimum RTO, EWMA parameters, and multiplier of variance RTT against different content popularity distribution gains. The obtained results are used to reconsider the Jacobson algorithm for accurately estimating RTO in CCN. Comparing the performance of the reconsidered Jacobson estimator with the existing solutions shows that it can estimate RTO simply and more accurately without any additional information or computation overhead.
... To overcome the above weaknesses of high-rate attacks, [1] pioneered a type of low-rate DoS attack named Shrew DoS attack. This attack exploits the homogeneity of the TCP's retransmission timeout (RTO) mechanism, i.e., in the modern Internet RTO is usually equal to its globally uniform lower bound minRTO [1], [3]. Consider a single TCP flow and a Shrew attack consisting of periodic " on-off " bursts. ...
... When the attacker launches a high rate burst that causes mass packet losses, the TCP flow has to enter timeout state and stop transmitting packet. This state will last for RTO (= minRTO) seconds [1], [3]. Afterwards, the victim flow enters slow start phase and begins to recover its transmission rate rapidly. ...
... 2 shows the general settings of the classical dumb-bell network used in this paper. Here, minRTO is set to its default value, 1 second [1], [3]. The TCP flows choose FTP as their application layer protocol, and they would fully consume the bottleneck bandwidth in the absence of attack. ...
Distributed Denial of Service (DDoS) attack has become one of the major threats to the Internet. Traditional brute-force, high-rate DDoS attacks expose many obvious anomaly features to defense systems, so that they can be easily detected and mitigated. In this paper we propose a new type of low-rate TCP-targeted DoS attack, called NewShrew, which exploits the deficiencies in TCP's timeout mechanism and slow start mechanism. This attack could significantly degrade TCP throughput, while evading the supervision of DoS prevention systems by inconspicuously consuming a small part of network capacity. We use theoretical analysis and numerical simulations to demonstrate the effectiveness of this attack for different RTT heterogeneity, TCP variant, and network environment. We reveal the interactions among the attack parameters, and the trade-offs between throughput degradation and attack cost. Moreover, we empirically show that NewShrew outperforms the classical Shrew DoS attack in terms of lower average attack rate (averagely 47.82%), higher attack efficiency (the ratio between throughput degradation inflicted by an attack and the average attack rate of the attack) with an average of 45.79%, and higher throughput degradation (averagely 11.54%) after deploying a typical defense mechanism (namely, RTO randomization). Our work innovatively exposes TCP slow start mechanism as a possible vulnerability to adversarial attacks, hence it opens new avenue to improving the resilience of TCP.
... Thus, estimating a good value for the retransmission timer not only involves estimating a property of the network path, but also a property of the remote connection peer. Third, if loss is due to congestion, it may behove the sender to wait longer than the maximum feedback time, in order to give congestion more time to drain from the network [2]. ...
... In [2], the parameters involved in RTO estimation (using Jacobson algorithm) are evaluated and the effect of their varying on estimator performance is studied. The results show that the performance of the estimators is dominated by their minimum values, and to a lesser extent, the timer granularity, while being virtually unaffected by how often round-trip time measurements are made or the settings of the parameters in the exponentially-weighted moving average of Jacobson algorithm. ...
... In [2], three ways of detecting spurious timeouts is proposed. Since we'll use them in our work, we explain them here. ...
Congestion control is a high priority and critical issue in today's networks. End-to-end congestion control mechanisms such as those in TCP are not enough to prevent congestion control and they must be supplemented by control mechanisms inside the network. In this paper we present a fuzzy logic approach for congestion control in TCP/IP networks. The proposed approach consists of three major parts: the Fuzzy Aggressive RTO Estimator (FARE), the Fuzzy Active Queue Management (Fuzzy AQM) and the Fuzzy Type 2 Scheduler. The FARE proposes an optimization of current RTO (retransmission TimeOut) estimation algorithm. By adding a detection mechanism for bad timeouts and undoing their side effects we are willing to have an aggressive estimator. By using a fuzzy system to adapt the K parameter of RTO estimation formula to network internal state, we try to answer the following questions: how much and when should our estimator act aggressively? On the other hand, the Fuzzy AQM computes the packet drop probability according to preconfigured fuzzy logic using the instantaneous queue length and number of packets dropped in a period of time as input variables. The Fuzzy Type 2 Scheduler uses the benefits of fuzzy type 2 controllers to adjust the service rate of the output buffers in the network routers. Our goal is to improve system performance and resource utilization in congested networks by a simple method. The simulation results with different web-like and FTP traffic show the superiority of TCP using FARE and Fuzzy AQM over a normal TCP. Furthermore it has been shown that the proposed Fuzzy Type 2 Scheduler has better performance than the traditional Weighted Round Robin (WRR) scheduler.
... In correspondence with the recommendation made by [3], many systems implement a minimum RTO of 1 second. It is this implementation that puts these systems at risk since a low-rate DoS attack can exploit it. ...
... Under heavy congestion, TCP reduces its congestion window size to 1 segment and sets the RTO to its minimum value. The recommended minimum value for the RTO is 1sec, as proposed by the study presented in [3]. This mechanism was chosen for dealing with cases of heavy congestion since it is the most conservative sender behavior. ...
... The important inter-burst period values are at 0.5 and 1 second. This is because they coincide with the RTO of 1 second that was proposed in [3]. ...
Low-rate TCP targeted denial of service attacks are a subset of DoS attacks that exploit the retransmission timeout (RTO) mechanism of TCP. In doing so, such an attack can drastically reduce throughput while producing little traffic relative to traditional DoS attacks. Since it produces only periodic traffic, a low-rate attack is difficult to detect and prevent. Another property of the periodic traffic, however, is that a low-rate attack's success depends on synchronization with the victim's RTO. A proposed defense to this attack is to randomize the RTO. In doing so, information can still be transmitted while the attacker is waiting and a connection will be able to avoid timing out successively. In this paper, we evaluate the effectiveness of randomizing the retransmission timeout in defending against low-rate TCP targeted denial of service attacks. Through experiments we show that such a defense can prevent a TCP flow from being throttled by a low-rate attack and still achieve respectable throughput. In addition, we will analyze the effectiveness of a low-rate DoS attack on the Linux implementation of TCP.
... One is as general information about the area of estimation of estimation of av_bw and second as reference for specific specialized consultation od basic concepts, functionality of estimation approaches, characteristics of the techniques developed and performance of certain tools. In [23], [24], [25], [26], [27], [28],, treats the basic concepts of the av_bw estimation area, such as capacity, availablebandwidth and the behavior of Internet traffic Self-similar and Burst traffic. Also authors in [29], [2], [17], broaden the previous basic concepts of the area of the estimation and measurement of av_bw. ...
... The av_bw of a link refers to the unused part of the total capacity of the link for a certain period of time. Therefore, although it appears that the capacity of a connection depends on the transmission rate of the technology used and the propagation medium used, it furthermore depends on the traffic load on that link that will vary with time [17], [27], [29]. ...
The estimation of the available bandwidth (av_bw) between two end nodes through the Internet, is an area that has motivated researchers around the world in the last twenty years, to have faster and more accurate tools; Due to the utility it has in various network applications; Such as routing management, intrusion detection systems and the performance of transport protocols. Different tools use different estimation techniques but generally only analyze the three most used metrics as av_bw, relative error and estimation time. This work expands the information regarding the evaluation literature of the current Available Bandwidth Estimation Tools (ABET's), where they analyze the estimation techniques, metrics, different generation tools of cross-traffic and evaluation testbed; Concentrating on the techniques and estimation methodologies used, as well as the challenges faced by open-source tools in high-performance networks of 10 Gbps or higher.
... 4.ACK Acknowledge [xi, 16, 17, 42–45, 55, 106] ACManager Admission Control Manager [84][85][86][87][88]API Application Programming Interface [84, 85] APP Application [10, 54, 55] AppAPI Application API [84][85][86] 88] AppID Application ID [85] ARP Address Resolution Protocol [37] ARQ Automatic Repeat-reQuest [16, 106] AS Autonomous System [14,[26][27][28][29]ASH Auxiliary Security Header [39] BPSK Binary Phase-Shift Keying [35] CAIDA Center for Applied Internet Data Analysis [20] xv xvi ...
... In [25] the authors observed that RTT is a poor approximation of the OWD and proposed a scheme that analytically derives the OWD, forward and reverse delay for asymmetric networks. Also, the analysis made in [26, 27] reveals that the Internet paths have large delay asymmetries, raising doubts about the accuracy of this method when used with real traffic. ...
... They are rare over all wireline paths [21], as well as on path's that include reliable wireless links that do not lose connectivity [14]. This is due to TCP's conservative retransmission timer [2], [15]. However, we believe that the problem will occur more frequently with the increasing number of hosts accessing the Internet via wide-area packet-radio networks. ...
... However, those segments are not logged by the packet filter until hiccup has terminated in second 42.6, when they get placed into the outbound interface buffer all at once. At that time, the sender has already performed one retransmission (marked as + in Figure 3) which was also queued by hiccup and can therefore only be seen in the receiver trace (see arrow (2) in Figure 3). The original transmission and the retransmission of that segment are the same point in the sender trace (see arrow (1) in Figure 3). ...
We propose an enhancement to TCP's error recovery scheme, which we call the Eifel algorithm . It eliminates the retransmission ambiguity, thereby solving the problems caused by spurious timeouts and spurious fast retransmits. It can be incrementally deployed as it is backwards compatible and does not change TCP's congestion control semantics. In environments where spurious retransmissions occur frequently, the algorithm can improve the end-to-end throughput by several tens of percent. An exact quantification is, however, highly dependent on the path characteristics over time. The Eifel algorithm finally makes TCP truly wireless-capable without the need for proxies between the end points. Another key novelty is that the Eifel algorithm provides for the implementation of a more optimistic retransmission timer because it reduces the penalty of a spurious timeout to a single (in the common case) spurious retransmission.
... The choice of T maxl and T maxr reflects a trade-off between bandwidth consumption and the accuracy of RTT measurements. Recent studies indicate that Internet RTTs resemble a heavy-tailed distribution with occasional spikes of extraordinarily high values [AP99]. Most of these studies focus on the accuracy of RTT estimation in the TCP protocol. ...
... Round trip time estimation in the presence of network dynamics is a complicated problem and has received much attention in the literature[Pax97b,Pax97a,AP99]. ...
... In prior work, a number of approaches have been proposed to estimate TCP's RTT. Trace-driven simulations reported in [6] to evaluate different RTT estimation algorithms show that the performance of the estimators is dominated by their minimum values and is not influenced by the RTT sample rate [6]. This last conclusion was challenged by the Eifel estimation mechanism [4], one of the most cited alternatives to TCP's original RTT estimator; Eifel can be used to estimate the RTT and set the RTO. ...
... In prior work, a number of approaches have been proposed to estimate TCP's RTT. Trace-driven simulations reported in [6] to evaluate different RTT estimation algorithms show that the performance of the estimators is dominated by their minimum values and is not influenced by the RTT sample rate [6]. This last conclusion was challenged by the Eifel estimation mechanism [4], one of the most cited alternatives to TCP's original RTT estimator; Eifel can be used to estimate the RTT and set the RTO. ...
In this paper, we explore a novel approach to end-to-end round-trip time (RTT) estimation using a machine-learning technique known as the experts framework . In our proposal, each of several ‘experts’ guesses a fixed value. The weighted average of these guesses estimates the RTT, with the weights updated after every RTT measurement based on the difference between the estimated and actual RTT.
Through extensive simulations, we show that the proposed machine-learning algorithm adapts very quickly to changes in the RTT. Our results show a considerable reduction in the number of retransmitted packets and an increase in goodput, especially in more heavily congested scenarios. We corroborate our results through ‘live’ experiments using an implementation of the proposed algorithm in the Linux kernel. These experiments confirm the higher RTT estimation accuracy of the machine learning approach which yields over 40% improvement when compared against both standard transmission control protocol (TCP) as well as the well known Eifel RTT estimator. To the best of our knowledge, our work is the first attempt to use on-line learning algorithms to predict network performance and, given the promising results reported here, creates the opportunity of applying on-line learning to estimate other important network variables.
... Allman and Paxson noted that an avail-bw estimate can give a more appropriate value for the ssthresh variable, improving the slow-start phase of TCP [2]. They recognized, however, the complexity of measuring avail-bw from the timing of TCP packets, and they focused instead on capacity estimates. ...
... The BDP is the product of the path's avail-bw with the connection's RTT. Previous efforts attempted to get a rough estimate of the BDP using capacity, rather than avail-bw, estimation techniques [16,2,38]. If the avail-bw is known, it may be possible to 'jump-start' a TCP connection from that rate (with appropriate pacing though), rather than using slow-start. ...
The available bandwidth (avail-bw) in a network path is of major importance in congestion control, streaming applications, QoS verification, server selection, and overlay networks. We describe an end-to-end methodology, called Self-Loading Periodic Streams (SLoPS), for measuring avail-bw. The basic idea in SLoPS is that the one-way delays of a periodic packet stream show an increasing trend when the stream's rate is higher than the avail-bw. We implemented SLoPS in a tool called pathload. The accuracy of the tool has been evaluated with both simulations and experiments over real-world Internet paths. Pathload is non-intrusive, meaning that it does not cause significant increases in the network utilization, delays, or losses. We used pathload to evaluate the variability ('dynamics') of the avail-bw in some paths that cross USA and Europe. The avail-bw becomes significantly more variable in heavily utilized paths, as well as in paths with limited capacity (probably due to a lower degree of statistical multiplexing). We finally examine the relation between avail-bw and TCP throughput. A persistent TCP connection can be used to roughly measure the avail-bw in a path, but TCP saturates the path, and increases significantly the path delays and jitter.
... Retransmission timeout (RTO) estimation is usually performed at the sender and plays an important role in acknowledgement-based reliable transport protocols such as TCP and SCTP. The goal is to achieve a compromise between balancing unnecessary retransmissions versus waiting too long to detect packet loss [6]. Real-time media applications typically use RTP over UDP for media transport. ...
... Such a scheduler is denoted using the notation [c-d]. Here between a and b packets are sent over path 1, followed by between c and d packets over path 2. For example a [2][3][4][6][7][8]scheduler will also distribute 30% of packets on path 1, and 70% on path 2, but in a bursty manner. ...
Delay-sensitive media applications typically prioritise timeliness over reliability, therefore preferring UDP over TCP. Retransmission is a method to compensate for packet loss and requires the receiver to provide timely feedback to the sender. Delaying the retransmission request too long may result in the retransmitted media arriving late. Alternatively, aggressive error estimation, where slightly delayed packets are seen as lost, results in unnecessary bandwidth usage and may contribute to further congestion of the network. We study receiver-based retransmission timeout (RTO) estimation in the context of real-time streaming over Multipath RTP and propose a solution in which we use statistical methods to provide accurate RTO prediction which allows for timely feedback. The proposed approach allows the receiver to accurately estimate the RTO when receiving media over multiple paths irrespective of the scheduling algorithm used at the sender. This enables a sender to take advantage of multiple paths for load balancing or bandwidth aggregation by scheduling media based on dynamic path characteristics.
... None of the above mentioned solutions help when all outstanding segments are lost or the only outstanding segment is lost, which can only be detected via a timeout. Several prior works [6, 20, 21, 10, 23] showed deficiencies in calculating the RTO and proposed improvements. The work in [21] investigated the effect of the large initial RTO whereas work in [6] showed that TCP's loss recovery performance is largely influenced by the minimum RTO value. ...
... Several prior works [6, 20, 21, 10, 23] showed deficiencies in calculating the RTO and proposed improvements. The work in [21] investigated the effect of the large initial RTO whereas work in [6] showed that TCP's loss recovery performance is largely influenced by the minimum RTO value. Spurious retransmission is, however, more likely to happen if a lower value for the minimum RTO is used. ...
Interactive applications do not require more bandwidth to go faster. Instead, they require less latency. Unfortunately, the current design of transport protocols such as TCP limits possible latency reductions. In this paper we evaluate and compare different loss recovery enhancements to fight tail loss latency. The two recently proposed mechanisms “RTO Restart” (RTOR) and “Tail Loss Probe” (TLP) as well as a new mechanism that applies the logic of RTOR to the TLP timer management (TLPR) are considered. The results show that the relative performance of RTOR and TLP when tail loss occurs is scenario dependent, but with TLP having potentially larger gains. The TLPR mechanism reaps the benefits of both approaches and in most scenarios it shows the best performance.
... Unnecessary reduction of cwnd caused by packet reordering leads to improper utilization of the link. 3. TCP ensures that the receiving application receives data in order. ...
... Increase the dupthresh by 1 for every detected false fast retransmission.2 Increase the dupthresh by the reordering length N for every detected false fast retransmission.3 We have provided these results in the Appendix since their code was not available during the entire phase of this thesis. ...
... Several authors [2,13,[22][23][24][25] reported that the most effective detection of congestion can occur in the gateway. It is in this context that the network-return techniques act. ...
In the absence of losses, TCP constantly increases the amount of data sent per instant of time. This behavior leads to problems that affect its performance, especially when multiple devices share the same gateway. Several studies have been done to mitigate such problems, but many of them require TCP side changes or a meticulous configuration. Some studies have shown promise, such as the use of gateway techniques to change the receiver’s advertised window of ACK segments based on the amount of memory in the gateway; in this work, we use the term “network-return” to refer to these techniques. In this paper, we present a new network-return technique called early window tailoring (EWT). For its use, it does not require any modification in the TCP implementations at the sides and does not require that all routers in the path use the same congestion control mechanism, and the use in the gateway is sufficient. With the use of the simulator ns-3 and following the recommendations of RFC 7928, the new approach was tested in multiple scenarios. The EWT was compared to drop-tail, RED, ARED, and the two network-return techniques—explicit window adaptation (EWA) and active window management (AWM). In the results, it was observed that EWT was shown to be efficient in congestion control. Its use avoided losses of segments, bringing expressive gains in the transfer latency and goodput and maintaining fairness between the flows. However, unlike other approaches, the most prominent feature of EWT is its ability to maintain a very high number of active flows at a given level of segment loss rate. The EWT allowed the existence of a number of flows, which is on average 49.3% better than its best competitor and 75.8% better when no AQM scheme was used.
... Hoe gives a refined PP method for guessing the available bandwidth in order to properly reset the ssthresh; the bandwidth is calculated by using the least-square estimation on the reception time of three ACKs corresponding to three closely-spaced packets [6]. Allman and Paxson evaluate the PP techniques and show that in practice they perform less well than expected [7]. Lai and Baker propose an evolution of the PP algorithm for measuring the link bandwidth in FIFO-queuing networks [8]. ...
Two widely known parameters of Transmission Control Protocol (TCP) used to control the flow of packets are Congestion Window (cwnd) & Slow Start Threshold (ssthresh). After congestion, slow start phase or fast-retransmit phase come in action wherein TCP has an important role in the reduction of these parameters. This is in response to packet loss identified by TCP. This in turn will cause unnecessary reduction of data flow & degradation of TCP throughput. Researchers have developed some algorithms to come out of this problem, WestwoodNR is one of them. WestwoodNR is using Bandwidth Estimation algorithm to estimate available bandwidth, to make effective use of available network capacity even after the congestion episode. It allows higher values of ssthresh & cwnd when it enters the fast-retransmit phase and slow start phase. In turn this algorithm claims better performance in terms of bandwidth utilization. The focus of this paper is on error recovery mechanisms suitable for WestwoodNR operating over the wireless sub path. These mechanisms have to address the increased bit error probability and temporary disruptions of wireless links. The efficiency of WestwoodNR within wireless scenarios is investigated and possible modifications that lead to higher performance are pointed out.
... A protocol is proposed in [18], for timing the acknowledgement of a retransmitted segment. If the acknowledgement returns in less than 3/4 × RTTmin, the retransmission is likely to be spurious. ...
In mobile ad hoc networks (MANETs), link failures and route changes occur most frequently, which may result in packet reordering. Transmission control protocol (TCP) performs poorly in such environment, which misinterprets the reordered packets as lost packets due to congestion. This has motivated us on developing a new protocol towards the packet reordering for improving the performance of TCP in MANETs. Optimal path or route selection is the major concern to improve the energy efficiency and network lifetime. In this paper, trust aware routing protocol for selecting optimal route in MANET is proposed. Based on this protocol, trust value for each node is calculated using direct and indirect trust value. Then the routing cost metric value is calculated and the path with minimum cost metric value is chosen as the best path in the network. After selecting the optimal path, data packet is to be transmitted through the optimal path. During the transmission, the data packet may get dropped or reordered due to congestion or mobility. A cross layer approach between network layer and transport layer to identify the dropped and reordered packets in the network is proposed in this paper. Simulation results are reported, which support this proposal.
... It is shown in experiment that an unnecessary TCP timeout results in loss of useful throughput, and TCP begins a new slow start. The fixed minimum RTO of one second was selected because it eliminated unnecessary timeouts [22]. In [23], a router based low rate TCP DoS detection approach is proposed. ...
... But if the RTO score is set to very high value then the impact on retransmitted packets will not happen but if the congestion happened then recovery from congestion will be delayed. The reader can refer to works by Allman and Paxson[32]for detailed study of TCP's timers. Let's take an example to understand TCP timer behaviors. ...
In this paper the authors have tried to measure the impact of different variants of pulsating distributed denial of service attacks on the self-similar nature of the network traffic and to see if the variation in the H index could be used for distinguishing them from normal network traffic.
... The last two are examined more deeply in [19] and [20] respectively, and they are considered as a starting point to build complex solution about RTT estimation. In [21], the authors analyze EWMA parameters in TCP retransmission timeout estimation. In [22] and [23], the authors consider techniques to monitor the One Way Delay in a passive way with minimal overhead. ...
Optimal interface selection is a key mobility management issue in heterogeneous wireless networks. Measuring the physical or link level performance on a given wireless access networks does not provide a reliable indication of the actually perceived level of service. It is therefore needed to take measurements at IP level, on the (bidirectional) paths from the Mobile Host to the node that is handling the mobility, over different heterogeneous networks. In this paper, we propose and analyze mechanisms for connectivity check and performance (network delay and packet loss) monitoring over IP access networks, combining active and passive monitoring techniques. We evaluate the accuracy and timeliness of the performance estimates and provide guidelines for tuning up the parameters. From the implementation perspective, we show that using application level measurements is highly CPU intensive, while a kernel based implementation has comparably a very low CPU usage. The Linux kernel implementation results in an efficient use of batteries in Mobile Hosts and intermediate Mobility Management Nodes can scale up to monitoring thousands of flows. The proposed solutions have been implemented in the context of a specific mobility management solution, but the results are of general applicability. The Linux implementation is available as Open Source.
... Allman and Paxson have found in [8] that RTOmin and the timer granularity are crucial for a good RTO estimation. They identified the clock granularity (500 ms for most OSs at that time) and the delayed acknowledgments (delayed ACKs, usually set to 200 ms) to be two of the main obstacles. ...
... Actually, the communication delay and throughput are two dependent variables, as delay spikes on 3G based Internet communication can cause spurious TCP timeouts leading to significant throughput degradation. The problem is that delay on Internet connections is highly variable resulting for instance from route flipping [17]. On the one hand, underestimation of round trip delay (RTT) leads to a premature retransmission timeout in case there is no loss or the retransmission could be handled by the fast retransmission mechanism. ...
Wide Area Monitoring Systems (WAMS) utilizing synchrophasor measurements is considered one of the essential parts in smart grids that enable system operators to monitor, operate, and control power systems in wide geographical area. On the other hand, high-speed, reliable and scalable data communication infrastructure is crucial in both construction and operation of WAMS. Universal mobile Telecommunication System (UMTS), the 3G standard for mobile communication networks, was developed to provide high speed data transmission with reliable service performance for mobile users. Therefore, UMTS is considered a promising solution for providing a communication infrastructure for WAMS. 3G based EWAMS (Egyptian wide area Monitoring System) is designed and implemented in Egypt through deployment a number of frequency disturbance recorders (FDRs) devices on a live 220kV/500kV Egyptian grid in cooperation with the Egyptian Electricity Transmission Company (EETC). The developed EWAMS can gather information from 11 FDRs devices which are geographically dispersed throughout the boundary of the Egyptian power grid and to a remote data management center located at Helwan University. The communication performance for the developed EWAMS in terms of communication time delay, throughput, and percentage of wasted bandwidth are studied in this paper. The results showed that the system can achieve successfully the communication requirements needed by various wide area monitoring applications.
... The existing solutions for application-limited flows can be categorized into two classes, one class of solutions triggers FR more often [84,[89][90][91] where the other class [92][93][94][95][96][97][98][99] improve the existing RTO mechanism so that less time is spent to detect losses. None of these existing solutions perform well when tail loss (last segment(s) in a flow is lost) happens. ...
Networking research and development have historically focused on increasing network throughput and path resource utilization, which particularly helped bulk applications such as file transfer and video streaming. Recent over-provisioning in the core of the Internet has facilitated the use of interactive applications like interactive web browsing, audio/video conferencing, multi- player online gaming and financial trading applications. Although the bulk applications rely on transferring data as fast as the network permits, interactive applications consume rather little bandwidth, depending instead on low latency. Recently, there has been an increasing concern in reducing latency in networking research, as the responsiveness of interactive applications directly influences the quality of experience.
To appreciate the significance of latency-sensitive applications for today's Internet, we need to understand their traffic pattern and quantify their prevalence. In this thesis, we quantify the proportion of potentially latency-sensitive traffic and its development over time. Next, we show that the flow start-up mechanism in the Internet is a major source of latency for a growing proportion of traffic, as network links get faster.
The loss recovery mechanism in the transport protocol is another major source of latency. To improve the performance of latency-sensitive applications, we propose and evaluate several modifications in TCP. We also investigate the possibility of prioritization at the transport layer to improve the loss recovery. The idea is to trade reliability for timeliness. We particularly examine the applicability of PR-SCTP with a focus on event logging. In our evaluation, the performance of PR-SCTP is largely influenced by small messages. We analyze the inefficiency in detail and propose several solutions. We particularly implement and evaluate one solution that utilizes the Non-Renegable Selective Acknowledgments (NR-SACKs) mechanism, which has been proposed for standardization in the IETF. According to the results, PR-SCTP with NR-SCAKs significantly improves the application performance in terms of low latency as compared to SCTP and TCP.
... Some applications benefit from knowing the bandwidth characteristics of networks [1] . For example , the congestion control of TCP determines the variable ssthresh based on an accurate estimation of available bandwidth , which greatly improves the throughput performance [2]. Network operator determines whether the capacity of network shall be upgraded according to the measurement of bandwidth utilization. ...
Available bandwidth estimation is an effective way to understand the situations of networks and applications. The packet pair technique analyzes the inter-arrival time of packet pair for the available bandwidth estimation in an end-to-end path. The inter-arrival time of packet pair is mostly caused by the queuing delay of narrow link which is sensitive to cross traffic. This paper proposes a new method using ACK pair probing (AProbing) to estimate end-to-end available bandwidth. It improves the packet pair technique with the probe gap model (PGM) which reduces the influences of cross traffic, and reconstructs the acknowledgements (ACKs) of Transmission Control Protocol (TCP) which reduces the overhead of measurement. AProbing has been implemented and verified in NS-3. The simulation results show that the accuracy of AProbing is similar with that of Pathload and 10% higher than the accuracies of Pathchirp and Cprobe. Moreover, AProbing significantly decreases the overhead of bandwidth estimation compared with these existed methods.
... When the forward and backward one-way delays along a roundtrip route are different, the one-way delays cannot be accurately estimated by halving the round-trip delays. The asymmetric forward and backward one-way delays are caused by two facts: the asymmetric round-trip routes in current Internet [5], and the asymmetric queuing delays on the two one-way paths even when a round-trip route is symmetric. This paper describes a novel method of estimating the variable delay component and the constant delay component along an one-way path without using GPS or NTP. ...
One-way delay monitoring is necessary for the time-sensitive media streaming. The conventional methods of one-way delay measurement require the clocks used at both the source and sink nodes of a path have to be synchronized. Clock synchronization can be performed either through GPS or NTP-based methods. Both approaches bear certain limitations. This paper describes a novel method of estimating the variable and constant components of one-way delays without requiring the local clocks at the end nodes to be synchronized. This method requires a sequence of probing packet batches to be sent from a source to a sink along a fixed one-way path. The timestamps of sending and receiving the packets at the source and the sink are recorded with respect to the local clocks. The variable and constant components of one-way delays can be estimated based on carefully crafting the relationship between the timestamps measured with respect to local clocks. The variable delay component is estimated based on the difference between the inter-arrival times between batches that are measured at the source and the sink, respectively. The constant delay component offset by the clock skew can then be estimated by making use of the estimated values of the variable delay component and the timestamps measured locally at the source and the sink. A novel concept of nominal service duration for a packet batch is introduced in modeling the one-way delays. The introduction of the nominal service duration makes this estimation method to be able to estimate the variable component of one-way delays in the existence of cross traffic. Without the knowledge of the offset of the local clocks at the source and the sink, this estimation method can only estimate the constant delay component up to the offset value of the actual constant delay.
... This implies that the number of accessible states is almost infinite. In addition the structure of the Internet is so complex that its 'State' cannot even be estimated accurately [4]. Therefore, unlike previous "multiservice" networks, the Internet cannot easily be managed using an information modelling approach based on finite-state-machine control theory. ...
Future multiservice networks will be multifractal, and thus impossible to control accurately using conventional finite-state-machine models. Active networking can reduce the cost to operators by delegating management to network users, but this will increase the overall system complexity. Management architectures for active networks must therefore use a new approach. An open-ended alternative to existing architectures is proposed, based on delivering homeostasis using environmental control and feedback hierarchies.
... Performance issues related to retransmission procedures, including alternatives to the Jacobson Algorithm, have been noted and addressed several times in the literature. Much work has been focused on late retransmission and other optimizations of the overall retransmission scenarios [12], [13]. Many authors approach this problem with a " holistic " or overall perspective on the retransmission procedures where RTT estimation contributes to triggering these procedures. ...
... Unlike TCP Reno, which simply halves the congestion window after three dupacks, TCPW attempts to make a more intelligent decision. It selects a slow-start threshold and a congestion window that are consistent with the effective connection rate at the time of congestion.These types of techniques for bandwidth estimation have been proposed before, (packet pair[31] and TCP Vegas[32]) but, due to technical reasons they have not been deployed onto the network. The key thing about TCPW is that it probes the network for the actual rate that a connection is achieving during the data transfer, not the available bandwidth before the connection is started. ...
... If the sender does not receive an Ack before a specified timeout value, then the timer expires and the data is considered lost. A detailed study of the effect of various timeout settings can be found in [20]. ...
The Internet is getting richer, and so the services. The richer the services, the more the users demand.
The more they demand, the more we guarantee(1).
This thesis investigates the congestion control mechanisms for interactive multimedia streaming
applications. We start by raising a question as to why the congestion control schemes are not widely
deployed in real-world applications, and study what options are available at present. We then discuss and
show some of the good reasonings that might have made the control mechanism, specifically speaking
the rate-based congestion control mechanism, not so attractive.
In an effort to address the problems, we identify the existing problems from which the rate-based
congestion control protocol cannot easily escape. We therefore propose a simple but novel windowbased
congestion control protocol that can retain smooth throughput property while being fair when
competing with TCP, yet still being responsive to the network changes.
Through the extensive ns-2 simulations and the real-world experiments, we evaluate TFWC, our
proposed mechanisms, and TFRC, the proposed IETF standard, in terms of network-oriented metrics
(fairness, smoothness, stability, and responsive), and end-user oriented metrics (PSNR and MOS) to
throughly study the protocol’s behaviors. We then discuss and conclude the options of the evaluated
protocols for the real application. (1)We as congestion control mechanisms in the Internet.
... Dado que está recomendado su uso con un valor estándar de RT O para todos los flujos[4]. ...
... In particular, the necessity for rather frequent measurements (e.g., per packet) is not well understood. There is some empirical evidence that such frequent sampling may not have a significant benefit [Allman99]. ...
We first study, at a microscopic level, the effects of various types of packet losses on TCP performance in a small data center. Then based on the findings we propose a simple recovery mechanism to combat the drawbacks of the long retransmission timeout. We emphasize through our empirical study that packet losses that occur at the tail of short-lived flows and/or bursty losses that span a large fraction of the congestion window are frequent in data center networks; and, in most cases, especially for short-lived flows, they result in a loss recovery that incurs waiting for a long retransmission timeout (RTO). The negative effect of frequent RTOs on the FCT is dramatic, yet recovery via RTO is merely a symptom of the pathological design of TCP's minimum RTO mechanism (set by default to the Internet scale). We propose the so-called Timely Retransmitted ACKs (T-RACKs), a very simple recovery mechanism for data centers, implemented as a shim layer between the virtual machines layer and the end-host NIC, to bridge the gap between TCP's huge RTO and the actual round trip times experienced in the data center. Compared to alternative solutions such as DCTCP, our TRACKS has the virtue of not requiring any modification to TCP, which makes it readily deployable in virtualized multi-tenant public data centers. Experimental results show considerable improvements in the FCT distribution.
Available bandwidth estimation is very important for network operators, users, and bandwidth-sensitive applications. In the last 20 years, various techniques and systems have been proposed to estimate end-to-end available bandwidth. They were mostly tested in simulation or inside small-scale networks, but they can’t consistently offer satisfying accuracy over the Internet. An active probing method SOProbe is proposed, and it measures end-to-end available bandwidth from only the installed host. The key idea of SOProbe is to identify the rate range where the available bandwidth resides. To archives this, SOProbe sends probe packets at selected transmission rates, and tries to catch the relationship between probe packets and available bandwidth.
TCP-targeted low-rate distributed denial-of-service (LDDoS) attacks pose a serious challenge to the reliability and security of the Internet. Among various proposed solutions, we are particularly interested in the Congestion Participation Rate (CPR) metric and the CPR-based approach. Through a simulation study, we show that the existing algorithm cannot simultaneously achieve high TCP throughput while under attack and good fairness performance for new legitimate TCP flows in normal times. We then propose a new version of the CPR-based approach to overcome the tradeoff. Simulation results show that it preserves TCP throughput while under attack fairly well, yet maintains fairness for new TCP flows in normal times.
In this article a framework is introduced that can be used to analyse the effects & requirements of P2P applications onapplication and on network layer. P2P applications are complex and deployed on a large scale, pure packet level simulations do not scale well enough to analyse P2P applications in a large network with thousands of peers. It is also difficult to assess the effect of application level behavior on the communication system. We therefore propose an approach starting with a more abstract and therefore scalable application level simulation. For the application layer a specific simulation framework wasdeveloped. The results of the application layer simulations plus some estimated background traffic are fed into a packet layer simulator like NS2 (or our lab testbed) in a second step to perform some detailed packet layer analysis such as loss and delay measurements. This can be done for a subnetwork of the original network to avoid scalability problems.
The restart method is based purely on the task completion time. In the previous chapter conditions under which the method is triggered have been investigated. When applying the restart method the only relevant question from a quantitative analysis point is when to restart. For the task under consideration its processing time is monitored and if the processing time exceeds a given value τ then the task is aborted and restarted from beginning. In stochastic terms one may say restart uses a completion time distribution, and a job is restarted when the remaining expected completion time is longer than the expected completion time when restarting the job, taking into account the time already waited for completion. This is similar to age replacement in preventive maintenance.
In this chapter we evaluate the performance of using TCP spoofing/splitting over satellite via a model-based study and a measurement-based study. Through analysis based on a few mathematical models we attempt to develop a basic understanding of the properties of TCP dynamics when connection spoofing is used, and further identify conditions under which using spoofing provides significant or marginal performance gain. Our measurement results are obtained from a commercial direct broadcast satellite system that uses TCP splitting at the network operating center (NOC). In particular these results show the performance of TCP splitting in combination with web caching. In both the model-based and the measurement-based studies we explore the effect of various factors including file size, congestion, and connection asymmetry. We use results and analysis from our models to better explain observations from the measurement. Finally we discuss the implications our findings and conclusions have on the design, deployment and provisioning of systems using TCP spoofing/splitting.
As discussed in Chaps. 1 and 3, clients performing synchronized reads across an increasing number of servers in high bandwidth, low latency data center environments, observe TCP’s throughput drop by one or two orders of magnitude below their link capacity. Labeled Incast, this pathological behavior of TCP is endured by a growing number of data center applications and services. Hence, a feasible solution that addresses the Incast problem is urgently needed. In this chapter, we provide a broad overview of existing Incast solutions followed by detailed description of our proposed techniques that are designed to address the Incast problem at the Transport Layer [1].
Measuring available bandwidth of an end-to-end network path has become a crucial operation in traffic engineering, congestion control, quality-of-service, and streaming applications. We present SOProbe, an active probing method of measuring the available bandwidth with only source host instead of intermediate routers or destination host on the targeted path. Equipped with SOProbe, one can unlimitedly measure available bandwidth of any path starting from his network nodes. The basic idea of SOProbe is to search for a narrow rate range in which the available bandwidth resides. To do this, SOProbe sends probing packets at different rates, and it monitors the dispersion of these packets on all links along the path to identify whether the available bandwidth is higher than the probing rates.
In the automotive sector, the product cycles are by means longer than in the mobile markets. This often leads to long timeframes for introducing innovations of the infotainment domain in automobiles. In this paper, we present an approach that allows to downsize on-board ECUs to support the typical case of running a specified number of applications while featuring the opportunity to back-off computational power in worst case scenarios. Furthermore, this concept enables car manufacturers to offer new functionality to customers over the complete product-lifetime without the need to upgrade the hardware of ECUs or to have cost intensive automotive high performance hardware available on-board. This is achieved by enabling automotive applications to utilize either built-in computational power, CE (Consumer Electronics) devices like the customer's smartphone or the computational power of cloud servers. Therefore, we extend OpenCL, an industry standard that is supported by many different companies, to work locally on automotive platforms as well as remotely over the network. This becomes possible with the emergence of IP-based in-car networks.
Low-rate denial of service (LDoS) is a new class of DoS attack, which exploits the deficiencies of the minimum RTO of TCP to send out attack packets about 10%-20% of normal traffic in short periodic pulses to a victim. It is hard to be detected through traditional detection mechanism. In this paper, an approach of detecting LDoS attack based on the model of small signal is proposed. The proposed approach takes statistics on the packets arriving in 30 seconds (sampling time is 10 ms, total of 3000 sampling points), and compares the statistical result with the characteristic judging value, which is settled as a threshold to indicate the difference between normal and attack flow. An eigenvalue-estimating matrix is established to estimate the attack period after LDoS attack being detected. Simulation results in NS-2 environment show that the proposed approach can detect the LDoS attack effectively.
WiFi is currently the most widely used communication standard over wireless LAN. Facing the trend that broad band and networking quality are emphasized, the IETF SIGTRAN working group and IEEE 802.11standards working groups proposed a new-generation communication protocol, stream control transmission protocol (SCTP) and wireless LAN communication standard, IEEE 802.11ac/VHT for transport layer and data-link layer. We wish the bandwidth of WLAN can reach Gbps level transmission performance to fulfill user's requirement towards service quality of wireless communication. Although SCTP and IEEE 802.11ac are emerging communication protocols designed to increase wireless communication quality, there have not been related researches conducting performance analysis on the two communication protocols. The results confirm that SCTP provides a better transmission performance than the traditional transmission control protocol (TCP). Furthermore, applying SCTP over IEEE 802.11ac results in a faster and more stable network environment. Consequently, SCTP and IEEE 802.11ac play key roles in realizing the vision of broadband transmissions in the digital convergence era.
Virtual machines (VMs) are widely adopted today to provide elastic computing services in datacenters, and they still heavily rely on TCP for congestion control. VM scheduling delays due to CPU sharing can cause frequent spurious retransmit timeouts (RTOs). Using current detection methods, we find that such spurious RTOs cannot be effectively identified because of the retransmission ambiguity caused by the delayed ACK (DelACK) mechanism. Disabling DelACK would add significant CPU overhead to the VMs and thus degrade the network’s performance. In this paper, we first report our practical experience about TCP’s reaction to VM scheduling delays. We then provide an analysis of the problem that has two components corresponding to VM preemption on the sender side and the receiver side, respectively. Finally, we propose PVTCP, a ParaVirtualized approach to counteract the distortion of congestion information caused by the hypervisor scheduler. PVTCP is completely embedded in the guest OS and requires no modification in the hypervisor. Taking incast congestion as an example, we evaluate our solution in a 21-node testbed. The results show that PVTCP has high adaptability in virtualized environments and deals satisfactorily with the throughput collapse problem.
To have an agriculture guaranteeing its good practices and thus, contributing to the environmental protection is a major issue of the sustainable development of rural areas. For that, it is necessary to be able to supervise the environment and to record the agricultural interventions. This monitoring requires to collect space-time data to supply the agrienvironmental information systems enable to build indicators allowing to take decisions, to provide controls... Data acquisition can be effective only if it is carried out in an automatic way through communication networks. Thus, within the framework of this work, the concept of agrienvironmental networks is presented. The characteristics and the constraints related to these kind of networks show that only the ad hoc wireless networks can be implemented to meet the needs. A state of the art of the ad hoc networks and the associated routings protocols shows that the creation of a new cooperative routing protocol is necessary in order to satisfy the agri-environmental network needs. This cooperative routing protocol is the principal contribution of this work. Then, the question of the Quality of Service in this kind of wireless networks is very important, in particular in a context of environment monitoring. Within the routing protocol suggested, the Quality of Service is approached according to several principles : estimate of the available bandwidth, improve of the network connectivity... The Quality of Service in the agri-environmental networks is thus another important point of this thesis work. The implementation of these principles is illustrated through various examples of agrienvironmental networks. This aspect is approached on the whole by the implementation of original modules as well on the hardware level as software. The proposed solutions allow to deploy fixed or mobile sensors, able to supervise the environment and thus, to understand better the man's intervention.
Most existing TCP variants cannot distinguish between different packet loss causes within MANETs. TCP was, mainly, developed to deal with network congestion errors. While within MANETs, there are packet loss causes other than congestion. Studying the behaviour of TCP in front of such losses, we notice that TCP doesn't have always the optimum behaviour as it reacts, in most cases, without considering the loss cause. This misbehaviour might cause network performance degradation and resources' waste. To overcome this problem, many LDAs have been designed. However, these LDAs were optimized for data networks where wireless link is only the last hop, meaning that they might be inadequate for MANETs. Also, the proposed LDAs deal only with losses due to wireless channel and/or congestion-induced errors. We show, in this paper, the importance of dealing with a third loss cause that is common in MANETs, which is link failure. We propose a new TCP variant that is called TCP-WELCOME. TCP-WELCOME can: (i) identify the loss cause by coupling loss and delay information, and (ii) trigger the appropriate packet loss recovery according to the identified loss cause. The performance evaluation, through both simulations and experimental tests, shows that TCP-WELCOME optimizes both energy consumption and achievable throughput. TCPWELCOME does not change the standard and can operate with existing TCP variants.
The paper proposes a model which helps to investigate the competitive aspect of real networks in quantitative terms. Through theoretical analysis and numerical simulations, it shows that the competitive model has the universality for a weighted network. The relation between parameters in the weighted network and the competitiveness in the competitive network is obtained by theoretical analysis. Based on the expression of the degree distribution of the competitive network, the strength and degree distributions of the weighted network can be calculated. The analytical solution reveals that the degree distribution of the weighted network is correlated with the increment and initial value of edge weights, which is verified by numerical simulations. Moreover, the evolving pattern of a clustering coefficient along with network parameters such as the size of a network, an updating coefficient, an initial weight and the competitiveness are obtained by further simulations.
Cloud datacenters are increasingly adopting virtual machines (VMs) to provide elastic cloud services, with TCP being prevalently used for congestion control. In virtualized datacenters, the delays from the hypervisor scheduler can heavily contaminate RTTs sensed by VM senders, preventing TCP from correctly learning the physical network condition. In this dissertation, my direction is to paravirtualize the transport-layer protocol in the guest OS, making it automatically tolerate the virtualized running environment. I then present a preliminary solution, PVTCP, to overcome the distorted congestion information caused by VM scheduling delays.
TCP, packet loss is the sole indicator of network congestion. Even so, TCP hosts are not explicitly notified of lost packets, but must rely on timeouts and duplicate acknowledgements (ACKs) to indicate loss. We believe that changes in one-way transit times (OTTs) could also be used to indicate congestion. Queues in routers build up before they overflow, resulting in increased OTTs. If all senders could directly measure changes in OTT and back off when the OTT increases past a threshold, congestion could be alleviated. In this work, we introduce a variant of TCP, called Sync-TCP, which uses the Global Positioning System (GPS) to gather a connection’s OTT data. We use Sync-TCP as a platform for investigating techniques for detecting and responding to changes in OTTs.
This paper analyzes the algorithm used for estimating retransmission timeouts in connection-oriented IP-based transport protocols, such as the Transmission Control Protocol (TCP) and the Stream Control Transmission Protocol (SCTP). The estimation algorithm uses historical values of the round-trip time to estimate future round-trip delays, and so creates a maximum waiting time before triggering retransmission attempts. The purpose of the analysis is to question / validate some of the fundamental assumptions used in the estimation algorithm. The conclusion of the analysis is that the algorithm is somewhat mismatched to the application area. Alternative algorithms are discussed, and potential modifications are presented.
Congestion control is a recognized problem in complex networks. We have discovered that the Department of Defense's Internet Protocol (IP), a pure datagram protocol, and Transmission Control Protocol (TCP), a transport layer protocol, when used together, are subject to unusual congestion problems caused by interactions between the transport and datagram layers. In particular, IP gateways are vulnerable to a phenomenon we call congestion collapse, especially when such gateways connect networks of widely different bandwidth. We have developed solutions that prevent congestion collapse.
These problems are not generally recognized because these protocols are used most often on networks built on top of ARPANET IMP technology. ARPANET IMP based networks traditionally have uniform bandwidth, identical switching nodes, and are sized with substantial excess capacity. This excess capacity, and the ability of the IMP system to throttle the transmissions of hosts has for most IP/TCP hosts and networks, been adequate to handle congestion. With the recent split of the ARPANET into two interconnected networks and the growth of other networks with differing properties connected to the ARPANET, however, reliance on the benign properties of the IMP system is no longer enough to allow hosts to communicate rapidly and reliably. Improved handling of congestion is now mandatory for successful network operation under load.
Ford Aerospace and Communications Corporation, and its parent company, Ford Motor Company, operate the only private IP/TCP long-haul network in existence today. This network connects six facilities (one in Michigan, two in California, one in Colorado, one in Texas, and one in England) some with extensive local networks. This net is cross-tied to the ARPANET but uses its own long-haul circuits; traffic between Ford facilities flows over private leased circuits, including a leased transatlantic satellite connection. All switching nodes are pure IP datagram switches with no node-to-node flow control, and all hosts run software either written or heavily modified by Ford or Ford Aerospace. Bandwidth of links in this network varies widely, from 1200 to 10,000,000 bits per second. In general, we have not been able to afford the luxury of excess long-haul bandwidth that the ARPANET possesses, and our long-haul links are heavily loaded during peak periods. Transit times of several seconds are thus common in our network.
Because of our pure datagram orientation, heavy loading, and wide variation in bandwidth, we have had to solve problems that the ARPANET/MILNET community is just beginning to recognize. Our network is sensitive to suboptimal behavior by host TCP implementations, both on and off our own net. We have devoted considerable effort to examining TCP behavior under various conditions, and have solved some widely prevalent problems with TCP. We present here two problems and their solutions. Many TCP implementations have these problems; if throughput is worse through an ARPANET/MILNET gateway for a given TCP implementation than throughput across a single net, there is a high probability that the TCP implementation has one or both of these problems .
Repeated observation of TCP retransmission timer problems stimulated investigation into the roles and limitations of timers. Timers are indispensable tools in building up reliable distributed systems. However, as the experience with the TCP retransmission timer has shown, timers have intrinsic limitations in offering optimal performance. Any timeout based action is a guess based on incomplete information, and as such is bound to be non-optimal. We conclude that, if we aim at high performance, we should use external events as a first line of defense against failures, and depend on timers only in cases where external notification has failed.
This paper presents a control-theoretic approach to reactive flow control in networks that do not reserve bandwidth. We assume a round-robin-like queue service discipline in the output queues of the network's switches, and propose deterministic and stochastic models for a single conversation in a network of such switches. These models motivate the Packet-Pair rate probing technique, and a provably stable rate-based flow control scheme. A Kalman state estimator is derived from discrete-time state space analysis, but there are difficulties in using the estimator in practice. These difficulties are overcome by a novel estimation scheme based on fuzzy logic. We then present a technique to extract and use additional information from the system to develop a continuous-time system model. This is used to design a variant of the control law that is also provably stable, and, in addition, takes control action as rapidly as possible. Finally, practical issues such as correcting parameter drift and coordination with window flow control are described.
The authors present random early detection (RED) gateways for
congestion avoidance in packet-switched networks. The gateway detects
incipient congestion by computing the average queue size. The gateway
could notify connections of congestion either by dropping packets
arriving at the gateway or by setting a bit in packet headers. When the
average queue size exceeds a present threshold, the gateway drops or
marks each arriving packet with a certain probability, where the exact
probability is a function of the average queue size. RED gateways keep
the average queue size low while allowing occasional bursts of packets
in the queue. During congestion, the probability that the gateway
notifies a particular connection to reduce its window is roughly
proportional to that connection's share of the bandwidth through the
gateway. RED gateways are designed to accompany a transport-layer
congestion control protocol such as TCP. The RED gateway has no bias
against bursty traffic and avoids the global synchronization of many
connections decreasing their window at the same time. Simulations of a
TCP/IP network are used to illustrate the performance of RED gateways
RFC 2001 [RFC2001] documents the following four intertwined TCP congestion control algorithms: Slow Start, Congestion Avoidance, Fast Retransmit, and Fast Recovery. RFC 2581 [RFC2581] explicitly allows certain modifications of these algorithms, including modifications that use the TCP Selective Acknowledgement (SACK) option [MMFR96], and modifications that respond to "partial acknowledgments" (ACKs which cover new data, but not all the data outstanding when loss was detected) in the absence of SACK. This document describes a specific algorithm for responding to partial acknowledgments, referred to as NewReno. This response to partial acknowledgments was first proposed by Janey Hoe in [Hoe95]. 1. Introduction For the typical implementation of the TCP Fast Recovery algorithm described in [RFC2581] (first implemented in the 1990 BSD Reno release, and referred to as the Reno algorithm in [FF96]), the TCP data sender only retransmits a packet after a retransmit timeout has occurred, or afte...
The behavior of the TCP protocol in simple situations is well-understood, but when multiple connections share a set of network resources the protocol can exhibit surprising phenomena. Earlier studies have identified several such phenomena, and have analyzed them using simulation or observation of contrived situations. This paper shows how, by analyzing traces of a busy segment of the Internet, it is possible to observe these phenomena in "real life" and measure both their frequency and their effects on performance. A TCP implementation might use similar techniques to support rate-based congestion control.
This paper uses simulations to explore the benefits of adding selective acknowledgments (SACK) and selective repeat to TCP. We compare Tahoe and Reno TCP, the two most common reference implementations for TCP, with two modified versions of Reno TCP. The first version is New-Reno TCP, a modified version of TCP without SACK that avoids some of Reno TCP's performance problems when multiple packets are dropped from a window of data. The second version is SACK TCP, a conservative extension of Reno TCP modified to use the SACK option being proposed in the Internet Engineering Task Force (IETF). We describe the congestion control algorithms in our simulated implementation of SACK TCP and show that while selective acknowledgments are not required to solve Reno TCP's performance problems when multiple packets are dropped, the absence of selective acknowledgments does impose limits to TCP's ultimate performance. In particular, we show that without selective acknowledgments, TCP implementations are constrained to either retransmit at most one dropped packet per round-trip time, or to retransmit packets that might have already been successfully delivered. 1
We use simulation to study the dynamics of the congestion control algorithm embedded in the BSD 4.3-Tahoe TCP implementation. We investigate the simple case of a few TCP connections, originating and terminating at the same pair of hosts, using a single bottleneck link. This work is an extension of our earlier work ([16]), where one-way traffic (i.e., all of the sources are on the same host and all of the destinations are on the other host) was studied. In this paper we investigate the dynamics that results from two-way traffic (in which there are data sources on both hosts). We find that the one-way traffic clustering and loss-synchronization phenomena discussed in [16] persist in this new situation, albeit in a slightly modified form. In addition, there are two new phenomena not present in the earlier study: (1) ACK-compression, which is due to the interaction of data and ACK packets and gives rise to rapid fluctuations in queue length, and (2) an out-of-phase queue-synchronization mo...
Made possible by progress in fiber-optic and VLSI technologies, networks offering increasing transmission capacity at decreasing
error rates are becoming available. New applications would benefit from this bandwidth but software protocol processing rates
have not kept up with available raw transmission speed. We present a comparative survey of techniques used at the transport
layer in eight representative protocols, most of which were designed to improve this situation. The protocols are the relevant
portions of the APPN, Datakit, Delta-t, NETBLT, OSIITP4, TCP, VMTP, and XTP architectures. We then go on to discuss which
of those techniques seem the most promising to us. An extensive list of references is included.
Modern implementations of TCP contain four intertwined algorithms that have never been fully documented as Internet standards: slow start, congestion avoidance, fast retransmit, and fast recovery. (2) and (3) provide some details on these algorithms, (4) provides examples of the algorithms in action, and (5) provides the source code for the 4.4BSD implementation. RFC 1122 requires that a TCP must implement slow start and congestion avoidance (Section 4.2.2.15 of (1)), citing (2) as the reference, but fast retransmit and fast recovery were implemented after RFC 1122. The purpose of this document is to document these four algorithms for the Internet. Acknowledgments
This paper is a brief description of (i) --(v) and the rationale behind them. (vi) is an algorithm recently developed by Phil Karn of Bell Communications Research, described in [15]. (vii) is described in a soon-to-be-published RFC (ARPANET "Request for Comments")
We have developed a Forward Acknowledgment (FACK) congestion control algorithm which addresses many of the performance problems recently observed in the Internet. The FACK algorithm is based on first principles of congestion control and is designed to be used with the proposed TCP SACK option. By decoupling congestion control from other algorithms such as data recovery, it attains more precise control over the data flow in the network. We introduce two additional algorithms to improve the behavior in specific situations. Through simulations we compare FACK to both Reno and Reno with SACK. Finally, we consider the potential performance and impact of FACK in the Internet.
We discuss findings from a large-scale study of Internet packet dynamics conducted by tracing 20000 TCP bulk transfers between 35 Internet sites. Because we traced each 100-kbyte transfer at both the sender and the receiver, the measurements allow us to distinguish between the end-to-end behavior due to the different directions of the Internet paths, which often exhibit asymmetries. We: (1) characterize the prevalence of unusual network events such as out-of-order delivery and packet replication; (2) discuss a robust receiver-based algorithm for estimating “bottleneck bandwidth” that addresses deficiencies discovered in techniques based on “packet pair;” (3) investigate patterns of packet loss, finding that loss events are not well modeled as independent and, furthermore, that the distribution of the duration of loss events exhibits infinite variance; and (4) analyze variations in packet transit delays as indicators of congestion periods, finding that congestion periods also span a wide range of time scales
The Internet is rapidly growing in number of users, traffic
levels, and topological complexity. At the same time it is increasingly
driven by economic competition. These developments render the
characterization of network usage and workloads more difficult, and yet
more critical. Few recent studies have been published reporting Internet
backbone traffic usage and characteristics. At MCI, we have implemented
a high-performance, low-cost monitoring system that can capture traffic
and perform analyses. We have deployed this monitoring tool on OC-3
trunks within the Internet MCI's backbone and also within the
NSF-sponsored vBNS. This article presents observations on the patterns
and characteristics of wide-area Internet traffic, as recorded by MCI's
OC-3 traffic monitors. We report on measurements from two OC-3 trunks in
MCI's commercial Internet backbone over two time ranges (24-hour and
7-day) in the presence of up to 240,000 flows. We reveal the
characteristics of the traffic in terms of packet sizes, flow duration,
volume, and percentage composition by protocol and application, as well
as patterns seen over the two time scales
A comparative survey is presented of techniques used at the
transport layer in eight representative protocols, most of which were
designed to improve the protocol processing rate. The protocols are the
relevant portions of the APPN, Datakit, Delta-t, NETBLT, OSI/TP4, TCP,
VMTP, and XTP architectures. The protocols are described, and the
functions under consideration are defined. No distinction is made as to
whether these functions are carried out in a LAN, MAN, or WAN
environment. The objective is to provide reliable, end-to-end
transmission of data. The mechanisms required to support connection
management, acknowledgements, flow control, and error handling are
examined. Suitable techniques for designing light-weight transport
protocols are identified. A discussion is presented as to which
technique seems the most promising
We discuss the problem of detecting errors in measurements of the total delay experienced by packets transmitted through a wide-area network. We assume that we have measurements of the transmission times of a group of packets sent from an originating host, A, and a corresponding set of measurements of their arrival times at their destination host, B, recorded by two separate clocks. We also assume that we have a similar series of measurements of packets sent from B to A (as might occur when recording a TCP connection) , but we do not assume that the clock at A is synchronized with the clock at B, nor that they run at the same frequency. We develop robust algorithms for detecting abrupt adjustments to either clock, and for estimating the relative skew between the clocks. By analyzing a large set of measurements of Internet TCP connections, we find that both clock adjustments and relative skew are sufficiently common that failing to detect them can lead to potentially large errors when ...
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 startup 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 Retransmi...
Accurately characterizing end-to-end Internet dynamics --- the performance that a user actually obtains from the lengthy series of network links that comprise a path through the Internet --- is exceptionally difficult, due to the network's immense heterogeneity. It can be impossible to gauge the generality of findings based on measurements of a handful of paths, yet logistically it has proven very difficult to obtain end-to-end measurements on larger scales. At the heart of our work is a "measurement framework" we devised in which a number of sites around the Internet host a specialized measurement service. By coordinating "probes" between pairs of these sites we can measure end-to-end behavior along O(N^2) paths for a framework consisting of N sites. Consequently, we obtain a superlinear scaling that ...
We describe tcpanaly, a tool for automatically analyzing a TCP implementation's behavior by inspecting packet traces of the TCP's activity. Doing so requires surmounting a number of hurdles, including detecting packet filter measurement errors, coping with ambiguities due to the distance between the measurement point and the TCP, and accommodating a surprisingly large range of behavior among different TCP implementations. We discuss why our efforts to develop a fully general tool failed, and detail a number of significant differences among 8 major TCP implementations, some of which, if ubiquitous, would devastate Internet performance. The most problematic TCPs were all independently written, suggesting that correct TCP implementation is fraught with difficulty. Consequently, it behooves the Internet community to develop testing programs and reference implementations. 1 Introduction There can be a world of difference between the behavior we expect of a transport protocol, and what we ...
This memo presents a set of TCP extensions to improve performance over large bandwidth*delay product paths and to provide reliable operation over very high-speed paths. It defines new TCP options for scaled windows and timestamps, which are designed to provide compatible interworking with TCP's that do not implement the extensions. The timestamps are used for two distinct mechanisms: RTTM (Round Trip Time Measurement) and PAWS (Protect Against Wrapped Sequences). Selective acknowledgments are not included in this memo. This memo combines and supersedes RFC-1072 and RFC-1185, adding additional clarification and more detailed specification. Appendix C summarizes the changes from the earlier RFCs. TABLE OF CONTENTS 1. Introduction ................................................. 2 2. TCP Window Scale Option ...................................... 8 3. RTTM -- Round-Trip Time Measurement .......................... 11 4. PAWS -- Protect Against Wrapped Sequence Numbers ............. 17 5. C...
- Mark Allman
- Vern Paxson
- W Richard Stevens
Mark Allman, Vern Paxson y W. Richard Stevens. TCP Congestion
Control, Abril 1999. RFC 2581.
Desearíamos agradecer, asimismo, a los revisores de SIGCOMM: Sally Floyd, Paul Mallasch y Craig Partridge por sus útiles comentarios sobre este artículo. Finalmente, la idea clave de que el receptor puede determinar los paquetes del emisor enviados recíprocamente
- Sally La Presente Publicación Se Ha Beneficiado Significativamente De Conversaciones Mantenidas Con
- Floyd Y Reiner
- Ludwig
La presente publicación se ha beneficiado significativamente de
conversaciones mantenidas con Sally Floyd y Reiner Ludwig.
Desearíamos agradecer, asimismo, a los revisores de SIGCOMM: Sally
Floyd, Paul Mallasch y Craig Partridge por sus útiles comentarios sobre
este artículo. Finalmente, la idea clave de que el receptor puede
determinar los paquetes del emisor enviados recíprocamente ( § 3.4)
corresponde a Venkat Rangan.
Modified TCP Congestion Avoidance Algorithm Email to the end2end-interest mailing list
- Van Jacobson
[Jac90]
Van Jacobson. Modified TCP Congestion Avoidance
Algorithm, April 1990. Email to the end2end-interest
mailing list. URL: ftp://ftp.ee.lbl.gov/email/
vanj.90apr30.txt.
- Mark Vern Paxson
- Scott Allman
- William Dawson
- Jim Fenner
- Griner
[PAD + 99] Vern Paxson, Mark Allman, Scott Dawson, William
Fenner, Jim Griner, Ian Heavens, Kevin Lahey, Jeff
Semke, and Bernie Volz. Known TCP Implementation
Problems, March 1999. RFC 2525.