Improving the performance of interactive TCP applications using service differentiation

Noureddine, Waël; Tobagi, Fouad A.

doi:10.1016/s1389-1286(02)00265-7

Cited by 24 publications

(6 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…High-Speed TCP was proposed by Sally Floyd as a sender-side alternative congestion control algorithm [4]. High-Speed TCP attempts to improve the performance of TCP connections with large congestion windows.…”

Section: High-speed Tcpmentioning

confidence: 99%

See 1 more Smart Citation

Evaluation of replacement protocols and modifications to TCP to enhance ASC Wide Area Network performance.

Romero

2004

View full text Add to dashboard Cite

Historically, TCP/IP has been the protocol suite used to transfer data throughout the Advanced Simulation and Computing (ASC) community. However, TCP was developed many years ago for an environment very different from the ASC Wide Area Network (WAN) of today. There have been numerous publications that hint of better performance if modifications were made to the TCP algorithms or a different protocol was used to transfer data across a high bandwidth, high delay WAN. Since Sandia National Laboratories wants to maximize the ASC WAN performance to support the Thor's Hammer supercomputer, there is strong interest in evaluating modifications to the TCP protocol and in evaluating alternatives to TCP, such as SCTP, to determine if they provide improved performance. Therefore, the goal of this project is to test, evaluate, compare, and report protocol technologies that enhance the performance of the ASC WAN.

show abstract

Section: High-speed Tcpmentioning

confidence: 99%

“…TCP was first standardized in 1980 and did not include any congestion control mechanisms [4]. TCP congestion control was created to allow TCP connections to recover from a lost packet.…”

Section: Introductionmentioning

confidence: 99%

Evaluation of replacement protocols and modifications to TCP to enhance ASC Wide Area Network performance.

Romero

2004

View full text Add to dashboard Cite

show abstract

“…For example, if a router is able to determine, for each flow, the number of bytes or packets already transmitted, the flows could be divided into different priority classes with the highest priority given to the flows with the least number of transmitted bytes or packets. Such implementations are discussed in [15,7,12]. Another method, based on the packet arrival rates within each flow, is described in [16].…”

Section: Introductionmentioning

confidence: 99%

Two-level processor-sharing scheduling disciplines

Aalto

Ayesta

Nyberg-Oksanen

2004

Proceedings of the Joint International Conference on Measurement and Modeling of Computer Systems

View full text Add to dashboard Cite

Inspired by several recent papers that focus on scheduling disciplines for network flows, we present a mean delay analysis of Multilevel Processor Sharing (MLPS) scheduling disciplines in the context of M/G/1 queues. Such disciplines have been proposed to model the effect of the differentiation between short and long TCP flows in the Internet. Under MLPS, jobs are classified into classes depending on their attained service. We consider scheduling disciplines where jobs within the same class are served either with Processor Sharing (PS) or Foreground Background (FB) policy, and the class that contains jobs with the smallest attained service is served first. It is known that the FB policy minimizes (maximizes) the mean delay when the hazard rate of the job size distribution is decreasing (increasing). Our analysis, based on pathwise and meanwise arguments of the unfinished truncated work, shows that Two-Level Processor Sharing (TLPS) disciplines, e.g., FB+PS and PS+PS, are better than PS scheduling when the hazard rate of the job size distribution is decreasing. If the hazard rate is increasing and bounded, we show that PS outperforms PS+PS and FB+PS. We further extend our analysis to study local optimality within a level of an MLPS scheduling discipline.

show abstract

“…When TCP was first standardized in 1980 the standard did not include any congestion control mechanisms [1]. TCP congestion control was created to allow TCP connections to recover from a lost packet more gracefully.…”

Section: Introductionmentioning

confidence: 99%

Evaluation of TCP congestion control algorithms.

Long¹

2003

View full text Add to dashboard Cite

Sandia, Los Alamos, and Lawrence Livermore National Laboratories currently deploy high speed, Wide Area Network links to permit remote access to their Supercomputer systems. The current TCP congestion algorithm does not take full advantage of high delay, large bandwidth environments. This report involves evaluating alternative TCP congestion algorithms and comparing them with the currently used congestion algorithm. The goal was to find if an alternative algorithm could provide higher throughput with minimal impact on existing network traffic. The alternative congestion algorithms used were Scalable TCP and HighSpeed TCP. Network lab experiments were run to record the performance of each algorithm under different network configurations. The network configurations used were back-to-back with no delay, back-to-back with a 30ms delay, and two-to-one with a 30ms delay. The performance of each algorithm was then compared to the existing TCP congestion algorithm to determine if an acceptable alternative had been found. Comparisons were made based on throughput, stability, and fairness.

show abstract

Improving the performance of interactive TCP applications using service differentiation

Cited by 24 publications

References 26 publications

Evaluation of replacement protocols and modifications to TCP to enhance ASC Wide Area Network performance.

Evaluation of replacement protocols and modifications to TCP to enhance ASC Wide Area Network performance.

Two-level processor-sharing scheduling disciplines

Evaluation of TCP congestion control algorithms.

Contact Info

Product

Resources

About