TCP behavior in networks with dynamic propagation delay

TCP transmission is mostly governed by Re–Transmission Timeout (RTO) values of each segment sent across the network. It is found that the fairness in the RTO is more need on effective management of the TCP. Most of the model calculates the RTO based on the Round Trip Time (RTT), which is the critical parameter in determining the behaviors of TCP. The RTO lacks in reflecting network foot prints. This paper presents algorithm which improves RTO, considering the network foot print for each segment sent across the network. Algorithm uses 3–way connection establishments RTT of TCP as the primary value to fine-tune the RTO. Using this RTT as base value and Advertise Window value of the receiver it calculates the rate of change of throughput as function of time to fix the RTO. The experiment has being simulated through OPNET.

Large, distributed High Energy Physics (HEP) collaborations, such as D0, CDF and US-CMS, depend on stable and robust network paths between major world research centres. The evolving emphasis on data and compute Grids increases the reliance on network performance. Fermilab's experimental groups and network support personnel identified a critical need for WAN monitoring to ensure the quality and efficient utilization of such network paths. This has led to the development of the Network Monitoring system we will present in this paper. The system evolved from the IEPM-BW project, started at SLAC three years ago. At Fermilab this system has developed into a fully functional infrastructure with bi-directional active network probes and path characterizations. It is based on the Iperf achievable throughput tool, Ping and Synack to test ICMP/TCP connectivity. It uses Pipechar and Traceroute to test, compare and report hop-by-hop network path characterization. It also measures real file transfer performance by BBFTP and GridFTP. The Monitoring system has an extensive web-interface and all the data is available through standalone SOAP web services or by a MonaLISA client. Also in this paper we will present a case study of network path asymmetry and abnormal performance between FNAL and SDSC, which was discovered and resolved by utilizing the Network Monitoring system.

A vertical handoff occurs when a multi-homed mobile host roams from a network to another heterogeneous network. Nevertheless, the network before and after a vertical handoff usually have drastic different characteristics. In this paper, we proposed an end-to-end based Vertical-handoff Aware TCP, called VA-TCP, which promptly adapts TCP behavior during a vertical handoff. When entering a new network, on the basis of packet-pair algorithm, VA-TCP dynamically estimates the new network bandwidth and round-trip time by using out-of-band ICMP messages. After that, by the estimated bandwidth and round-trip time, VA-TCP adjusts the TCP parameters including congestion window size, slow-start threshold, retransmission timer and RTTs to respond to the new network environment. From the experimental result, the TCP throughput by the proposed VA-TCP scheme is around 2 times that by the original TCP.

Capacity estimation effective in an IP network path is a key enabling technology for proactive and reactive resource management as well as intelligent planning in next generation IP networks. We analyze capacity measurements at different Internet destinations. Particularly, we use the variable packet size (VPS) method to measure the per-hop capacity in some network paths. This method is based on round trip time (RTT) measurements as well as on the Internet control message protocol, ICMP. The RTT delay is approximated by a gamma distribution. We observe how the measurement accuracy of this active probing is affected by factors such as layer 2 device latencies, limited clock resolution at the probing host, or queuing delay. These factors cause non linearity on the RTT curves obtained and, consequently, inaccuracies on the estimation of path capacities. The use of robust regression to improve linear approximation in the VPS method is proposed. As a result, we obtain better qualitative and quantitative results.

Research into a transport protocol for high-speed networks is one of the most progressive areas in the field of computer communications. In high-speed networks using optical fiber as a transmission medium that gives the bandwidth from a hundred Mbits/s to Gbits/s, the roundtrip (propagation) delay and bandwidth product becomes large. It is well known that for an increase in this product, the throughput performance of the conventional error control procedures such as the go-back-N-based-ARQ scheme does degrade significantly. Therefore, a new protocol based on a selective-repeat ARQ scheme must be developed. The authors propose the stop-and-selective repeat (SSR) ARQ protocol and analyze the throughput performance of it. The throughput formula is exactly derived in terms of the roundtrip delay, the frame error rate, and the window size of from one to two roundtrip delays. The window size needed to offer the adequate throughput greater than 95% is shown.

Assigning jobs to optimum resources in a grid environment is the main aim of a grid scheduler. Communication cost has always been an important issue in grid environments. Proposing new scheduling algorithms to consider this cost accurately and allocate jobs to an optimum resource efficiently has always been of great importance. In this paper we have proposed a grid scheduling algorithm that is aware of the costs for different network paths. We have also considered soft real-time characteristics of jobs in our proposed algorithm. We have simulated and compared our algorithm with some key scheduling algorithms such as Least Load First (LLF), Random, and First Come First Served (FCFS) in Gridsim by considering soft real-time jobs. Results show the superiority of our scheduling algorithm due to its ability to predict network cost and to satisfy the demands of soft real-time tasks.

<div class="htmlview paragraph">Transceiver propagation (also known as wrap-around, loop-back, or round-trip) delay time and output voltage slew rates have a profound impact upon bit timing. Excessive propagation delay produces an error in the generated symbol time that may cause a bit timing error in a receiving node. Any difference between the compensation time of the protocol handler and the transceiver propagation delay will be seen in the transmitted pulse time. The larger the gap, the longer the transmitted bit time. The bit distortion increases even further if the receiving node is operating in a ground offset voltage with respect to the transmitter. Bit distortion is exacerbated as well when the output voltage slew rates are too slow. When the slew rates are too slow, the receiving bit time may be too short in a positive ground offset voltage condition and too long in a negative ground offset voltage condition. Either of these issues may also contribute to the inability of the protocol handler to successfully arbitrate the bus. The transmitting node may diagnose a short-to-ground condition and cause glitches on the bus.</div> <div class="htmlview paragraph">Transceiver propagation delay must be designed to be compatible with the compensation capabilities of the intended protocol handler(s). Output voltage slew rates must also be understood so that the total contribution of these two errors does not result in the inability to transmit successfully into maximum ground offset conditions.</div>

This paper studies the impact of variable transmission delays on the Transmission Control Protocol (TCP). Sudden delay variations, which are not uncommon in mobile networks, may degrade the performance since they may cause spurious TCP timeouts. The most important parameter in this context is the TCP retransmission timer. In this paper, we analyze TCP’s round-trip time estimation for bulk data traffic over wireless links. The main contribution is a new analytical model that accurately predicts the timeout duration from given network parameters. As a first result, the model shows that the round-trip time sampling rate has a significant impact on the timer characteristics. Therefore, the standardized estimation algorithm does not harmonize well with timestamp-based measurement. Second, we quantify the risk of spurious TCP timeouts triggered by changing round-trip times, in particular long off periods. We conclude that delay variations are only critical when they are on the order of seconds.

Designed for long propagation delay and frequent channel disruptions, delay tolerant networking (DTN) protocols are supporting the latest proposed optical space internetworking communication missions. While previous DTN-based testbeds experimented with bandwidth capacity in megabits, this paper presents a channel emulation testbed that supports multiple gigabit-bandwidth Ethernet clients and servers simultaneously. In addition to capacity, this testbed modeled optical link properties with channel emulation, and provided high flexibility by utilizing virtual LANs and link aggregation connected through a gigabit switch. The testbed emulated propagation delay, asynchronous channel rates, and bit errors, including free-space optical bit errors that varied as a function of time. We emulated an optical flight terminal relaying data between ground stations, and measured the maximum goodput of various DTN configurations while increasing the relay's round trip times. Results showed that for a 200 ms to one second round trip time, BP/TCPCL/TCP transmitted higher goodput than BP/LTP/UDP, and TCP/IP. For round trip times longer than one second, BP/LTP/UDP transmitted higher goodput than BP/TCPCL/TCP, and TCP/IP.

We propose the first practical solution to the longstanding problem of secure wireless routing in the presence of colluding attackers. Our secure routing protocol, Sprout, continuously tries new routes to the destination. Routes are probabilistically generated, with complete disregard for performance metrics. This makes Sprout uniquely resilient to attack: it cannot be tempted by shortcuts. In order to avoid compromised routes, and to ensure good overall performance, the quality of each active route is monitored by means of signed end-to-end acknowledgments. The amount of traffic sent on each route is adjusted accordingly. Sprout effectively mitigates the vast majority of known routing layer attacks, even when under assault from a large number of colluding attackers. Experiments on our 31-node testbed demonstrates the real-world performance of Sprout in terms of packet delivery ratio, round-trip times and TCP throughput. Our security analysis and simulation results show that Sprout is able to quickly find working paths in networks of hundreds of nodes and dozens or more attackers. For example, in a network of 200 nodes and an astounding 64 attackers, Sprout, on average, found a successful route within less than 10 attempts. Yet, in benign settings, Sprout provides TCP throughput within 15% of the shortest path throughput Overall, Sprout consistently delivers high, reliable performance in benign as well as hostile environments.

Grids offer the potential to carry out difficult computing tasks and achieve superior aggregate performance. However, grids are highly complex systems. They consist of heterogeneous resources on disparate hosts from various virtual organizations interconnected via a mixture of communication standards. Monitoring grid resources allows grid schedulers to adapt to changes in the status of these remote resources and the network paths between them. This is crucial to ensuring optimum performance. In this paper we introduce a distributed solution, called GridMAP, to collect network and end-host resource measurements, analyze their performance and feed these statistics and predictions back to schedulers. At this stage, we present our implementation of a passive TCP-SYN-based technique to provide GridMAP with round trip time and throughput measurements and we evaluate our approach against ping and iperf.

Satellite communication has become important in the current global data communication due to its large coverage area. Satellite links are affected by two major concerns: long propagation delays, and relatively high error rates. Standard TCP congestion control variants were not designed to address these challenges. They generally depend on round trip times (RTTs), and hence suffer heavily if the RTTs are large. Various TCP variants like Hybla have been proposed in order to optimize the channel throughput in heterogeneous networks incorporating satellite links. The error rate of satellite channels can be substantially mitigated using forward error correction (FEC) techniques. In such a situation, according to the results obtained by our comparative study, the performance of Cubic, which is designed for high-speed networks, is better than Hybla, in conditions of high-latency and low error rates.

The General Packet Radio Service (GPRS) is being deployed by GSM network operators world-wide, and promises to provide users with “always-on” data access at bandwidths comparable to that of conventional fixed-wire telephone modems. However, many users have found the reality to be rather different, experiencing very disappointing performance when, for example, browsing the web over GPRS. In this paper, we examine the causes, and show how unfortunate interactions between the GPRS link characteristics and TCP/IP protocols lead to poor performance. A performance characterization of the GPRS link-layer is presented, determined through extensive measurements taken over production networks. We present measurements of packet loss rates, bandwidth availability, link stability, and roundtrip time. The effect these characteristics have on TCP behavior are examined, demonstrating how they can result in poor link utilization, excessive packet queueing, and slow recovery from packet losses. Further, we show that the HTTP protocol can compound these issues, leading to dire WWW performance. We go on to show how the use of a transparent proxy interposed near the wired-wireless border can be used to alleviate many of these performance issues without requiring changes to either client or server end systems.

TCP is the standard transport protocol for reliable data communication over the internet. However, TCP behaviour over wireless mesh networks encounters random losses due to channel errors, changing topology and node distance. This paper aims to perform a comparison of TCP Old Tahoe and Tahoe in wireless mesh networks and analyze the throughput, propagation delay, packets retransmitted and number of duplicate acknowledgements received for varying node densities. The anomalous behaviour of TCP in the wireless network is documented, and probable explanations are provided. A comparison of the most suitable variant for each case is concluded. Overall, Tahoe shows 52% lower average throughput and 14% lower delay as compared to Old Tahoe. TCP Old Tahoe appears to be more suitable for lower density networks, whereas TCP Tahoe gives better results for higher density networks.

In this paper, we examine routing protocols with the shortest path in sensor networks. In doing this, we propose a genetic algorithm (GA)-based Ad Hoc On-Demand Multipath Distance Vector routing protocol (GA-AOMDV). We utilize a fitness function that optimizes routes based on the energy consumption in their nodes. We compare this algorithm with other existing ad hoc routing protocols including LEACH-GA, GA-AODV, AODV, DSR, EPAR, EBAR_BFS. Results prove that our protocol enhances the network performance in terms of packet delivery ratio, throughput, round trip time and energy consumption. GA-AOMDV protocol achieves average gain that is 7 to 22% over other protocols. Therefore, our protocol extends the network lifetime for data communications.