DPS queues with stationary ergodic service times and the performance of TCP in overload

Altman, Eitan; Jiménez, Tania; Kofman, Daniel

doi:10.1109/infcom.2004.1356984

Cited by 40 publications

(45 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [17] the authors consider non symmetric RTTs. They develop a DPS (Discriminatory Processor Sharing) queue model for TPC in the overload regime and test it with simulations in ns2.…”

Section: Other Modelsmentioning

confidence: 99%

“…Indeed, various papers in which sharing bandwidth is modeled by processor sharing report differences between the expected theoretical value and the simulation value [13,14,11,17] that vary from around 10% in [11] and go up to a factor of 10 in some situations in [13]. When such deviations occur, it is important to know whether they can be due to the imprecisions in the simulation or to real phenomena.…”

Section: Simulation Involving Flows With Heavy Tailed Size Distributionmentioning

confidence: 99%

“…Several research groups have examined its validity through simulations [17,14,11,13]. The conclusions of the papers have not always been encouraging.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Simulating Flow Level Bandwidth Sharing with Pareto distributed File Sizes

Rojas-Mora¹,

Jiménez²,

Altman³

2011

Proceedings of the 5th International ICST Conference on Performance Evaluation Methodologies and Tools

View full text Add to dashboard Cite

Our goal is to achieve deeper understanding of the limitation of queueing models, on one hand, and of common simulation practice, on the other hand, as tools for predicting performance of bandwidth sharing between competing TCP flows. In particular, we (i) present an overview of simulation problems that are expected to arise due to the very heavy tail of the distribution of the size of TCP flows, and (ii) through simulations we show that the average sojourn time of competing flows are quite sensitive to various network parameters. The understanding that we get from the first point allows us to better assess when are the conclusions from simulation results on bandwidth sharing reliable.Using simulations in ns2, we study bandwidth sharing under various load factors and show the benefit of using bootstrap as a post simulation tool for analyzing the simulation results.

show abstract

“…In [17] the authors consider non symmetric RTTs. They develop a DPS (Discriminatory Processor Sharing) queue model for TPC in the overload regime and test it with simulations in ns2.…”

Section: Other Modelsmentioning

confidence: 99%

Section: Simulation Involving Flows With Heavy Tailed Size Distributionmentioning

confidence: 99%

See 1 more Smart Citation

Simulating Flow Level Bandwidth Sharing with Pareto distributed File Sizes

Rojas-Mora¹,

Jiménez²,

Altman³

2011

Proceedings of the 5th International ICST Conference on Performance Evaluation Methodologies and Tools

View full text Add to dashboard Cite

show abstract

“…For instance, TCP flows that share a common bottleneck link but traverse heterogeneous routes, may experience diverse packet loss rates and round-trip delays. Because of TCP mechanics, these differences result in a significant discrepancy in the bandwidth shares, see for instance [2].…”

Section: Introductionmentioning

confidence: 99%

Differentiated bandwidth sharing with disparate flow sizes

Kessel

Núñez-Queija

Borst

Proceedings IEEE 24th Annual Joint Conference of the IEEE Computer and Communications Societies.

View full text Add to dashboard Cite

Abstract-We consider a multi-class queueing system operating under the Discriminatory Processor-Sharing (DPS) discipline. The DPS discipline provides a natural approach for modeling the flow-level performance of differentiated bandwidth-sharing mechanisms. Motivated by the extreme diversity in flow sizes observed in the Internet, we examine the system performance in an asymptotic regime where the flow dynamics of the various classes occur on separate time scales. Specifically, from the perspective of a given class, the arrival and service completions of some of the competing classes (called mice) evolve on an extremely fast time scale. In contrast, the flow dynamics of the remaining classes (referred to as elephants) occur on a comparatively slow time scale. Assuming a strict separation of time scales, we obtain simple explicit expressions for various performance measures of interest, such as the distribution of the numbers of flows, mean delays, and flow throughputs. In particular, the latter performance measures are insensitive, in the sense that they only depend on the service requirement distributions through their first moments. Numerical experiments show that the limiting results provide remarkably accurate approximations in certain cases 1 .

show abstract

“…[5]). An example of performance measures that depend on the whole distribution of service times but is insensitive to correlations is the growth rate of number of customers or of sojourn time in a (discriminatory) processor sharing queue in overload [4], [13].…”

Section: Introductionmentioning

confidence: 99%

On stochastic recursive equations and infinite server queues

Altman

Proceedings IEEE 24th Annual Joint Conference of the IEEE Computer and Communications Societies.

Self Cite

View full text Add to dashboard Cite

Abstract-The purpose of this paper is to investigate some performance measures of the discrete time G/G/ queue under a general arrival process. We assume more precisely that at each time unit a batch with a random size may arrive, where the sequence of batch sizes need not be i.i.d. All we request is that it would be stationary ergodic and that the service duration has a phase type distribution. Our goal is to obtain explicit expressions for the first two moments of number of customers in steady state. We obtain this by computing the first two moments of some generic stochastic recursive equations that our system satisfies. We then show that these class of recursive equations allow to solve not only thequeue but also a network of such queues. We finally investigate the process of residual activity time in a ¡ £ ¢ © ¡ £ ¢ queue under general stationary ergodic assumptions, obtain the unique stationary solution and establish coupling convergence to it from any initial state.

show abstract

DPS queues with stationary ergodic service times and the performance of TCP in overload

Cited by 40 publications

References 9 publications

Simulating Flow Level Bandwidth Sharing with Pareto distributed File Sizes

Simulating Flow Level Bandwidth Sharing with Pareto distributed File Sizes

Differentiated bandwidth sharing with disparate flow sizes

On stochastic recursive equations and infinite server queues

Contact Info

Product

Resources

About