Optimizing TCP Retransmission Timeout

Kesselman, Alex; Mansour, Yishay

doi:10.1007/978-3-540-31957-3_17

Cited by 30 publications

(16 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Several techniques have been designed to (i) detect and rectify the adverse impact of spurious retransmissions in an ongoing TCP transfer [17], [18], [19], [20], [21], and (ii) detect losses using alternate mechanisms [22], [23], [24], [25], [26], [27]. Unfortunately, due to deployment hurdles, most of these techniques have not been widely deployed in TCP implementations.…”

Section: Related Workmentioning

confidence: 99%

A Performance Study of Loss Detection/Recovery in Real-world TCP Implementations

Rewaskar

Kaur

Smith

2007

2007 IEEE International Conference on Network Protocols

View full text Add to dashboard Cite

Abstract-TCP is the dominant transport protocol used in the Internet and its performance fundamentally governs the performance of Internet applications. It is well-known that packet losses can adversely affect the connection duration of TCP connections-however, what is not fully understood is how well does the TCP design deal with losses. In this paper, we systematically evaluate the impact of design parameters associated with TCP's loss detection/recovery mechanisms on the performance of real-world TCP connections. For this, we rely on an analysis tool that partially emulates the sender-side TCP implementations of 5 prominent OSes for passively analyzing the traces of TCP connections. Our study conducts passive analysis of more than ¢ ¤ £¥ million real Internet TCP connections. We find that the recommended as well as widely-implemented settings of TCP parameters are not optimal for a significant fraction of Internet connections.

show abstract

Section: Related Workmentioning

confidence: 99%

A Performance Study of Loss Detection/Recovery in Real-world TCP Implementations

Rewaskar

Kaur

Smith

2007

2007 IEEE International Conference on Network Protocols

View full text Add to dashboard Cite

show abstract

“…Since the data packets are delayed and not lost, the resulting retransmission is unnecessary and the timeout is spurious. A study was conducted in [13] to quantify the sensitivity of TCP to sudden delay variation in mobile networks and it was shown in [8], [16] that optimal RTO value should not depend on just RTT measurements but other factors as well such as the TCP window size.…”

Section: A Delay Spikes and Spurious Timeoutsmentioning

confidence: 99%

Resource Efficiency in MANETs: Effect of Spurious Timeouts and Routing Protocol Dynamics

Mbarushimana

Shahrabi

2008

2008 IEEE 68th Vehicular Technology Conference

View full text Add to dashboard Cite

The presence of delay-sensitive traffic in QoSaware MANETs (Mobile Ad-hoc Networks) results into increased TCP spurious timeouts, thus wasting their scarce bandwidth. Furthermore, the different dynamics used by a routing protocol in discovering and maintaining routes have a significant impact on TCP and MANETs in general. In this paper, we investigate the combined effect of contentioninduced spurious timeouts and the routing protocol dynamics on the resource efficiency of TCP in the presence of prioritised VoIP traffic. Through simulation study, TCP performance is contrasted against that of our proposed enhancement to limit spurious timeouts in QoS-aware MANETs (RE-TCP). Our simulations results show that although proactive protocols deliver TCP traffic with the smallest delay and retransmit fewer TCP segments, reactive protocols are able to achieve better throughput and reduce TCP starvation in presence of high priority VoIP traffic. It has also been shown that RE-TCP is able to achieve a 10% gain in bandwidth utilization.

show abstract

“…Thus, to tackle this complexity, we introduced a probabilistic model of the latency imposed to single jobs in [1]. A probabilistic approach provides a black-box model which as been successfully applied in many scientific areas to model complex systems [3], [4], [5], [6]. In this paper, we propose to model the makespan of workflowbased applications which are representative of a large class of scientific grid applications.…”

Section: A Probabilistic Modelingmentioning

confidence: 99%

A Probabilistic Model to Analyse Workflow Performance on Production Grids

Glatard

Montagnat²,

Pennec³

2008

2008 Eighth IEEE International Symposium on Cluster Computing and the Grid (CCGRID)

View full text Add to dashboard Cite

Production grids are complex and highly variable systems whose behavior is not well understood and difficult to anticipate. The goal of this study is to estimate the impact of the variability of those infrastructures on the performance of workflow-based applications. A probabilistic model of workflows execution time is proposed and evaluated. Results show that the variability of the EGEE grid infrastructure impacts the execution time of a particular medical image analysis application by a factor 2. The model gives interesting insights on the grid behavior for different application parallelization modes. I. PERFORMANCE ANALYSIS ON PRODUCTION GRIDSIn many scientific areas, applications with stringent requirements for high performance computing, large data sets analysis and complex computation flows have emerged. Pushed by these new computational challenges very large scale production grids infrastructures have been deployed world-wide. Such widely distributed systems have been operating 24/7 over several years now, providing a sustained high end computing facility that many applications exploit routinely. The experience gained exploiting these systems shows that they can hardly be compared to traditional clusters performing on local area networks. For instance, we showed in a previous work that setting a timeout value to the jobs is mandatory on production grids whereas it is useless on most clusters [1]. Such differences may come from various factors. First, the reliability and homogeneity of clusters and local networks cannot be assumed on grids. Second, grids face very variable load patterns and race conditions originating from the shared exploitation by large user communities. Finally, the heterogeneity and the volatility of grid resources further increases the variability.Consequently, production grids exhibit hard to predict behaviors that result in variable overheads imposed to the computations from the users point of view. For instance, we observed that over thousands of computation tasks submitted to the EGEE production grid 1 in the same experimental conditions during months, an average delay of approximately 5 minutes with a standard deviation of the same order of magnitude (5 minutes) is experienced. For grid applications requiring the submission of a very large number of short (less than 1 hour long) jobs in parallel, such overheads are far from being negligible. As a result, applications computation time (makespans) 1 Now affiliated to the University of Amsterdam 2 Enabling Grids for E-sciencE, http://www.eu-egee.org are hardly forecastable, which makes performance analysis on production grids very difficult. In particular, the impact of the variability of the platform on the application should be quantified, as some works already suggested that it may have a strong negative impact on the applications [2].The objective of this paper is to propose a grid application makespan model that (i) aims at explaining the performance of applications on production grids, (ii) allows to study the impact of grid va...

show abstract

Optimizing TCP Retransmission Timeout

Cited by 30 publications

References 24 publications

A Performance Study of Loss Detection/Recovery in Real-world TCP Implementations

A Performance Study of Loss Detection/Recovery in Real-world TCP Implementations

Resource Efficiency in MANETs: Effect of Spurious Timeouts and Routing Protocol Dynamics

A Probabilistic Model to Analyse Workflow Performance on Production Grids

Contact Info

Product

Resources

About