Generalized stochastic performance models for loss-based congestion control

  • Authors:

  • Michele C. Weigle;Li Cheng;Jasleen Kaur;V. Kulkarni

  • Affiliations:

  • Department of Computer Science, Old Dominion University, Norfolk, VA 23529, USA;Department of Statistics and Operations Research, University of North Carolina, Chapel Hill, NC 27599, USA;Department of Computer Science, University of North Carolina, Chapel Hill, NC 27599, USA;Department of Statistics and Operations Research, University of North Carolina, Chapel Hill, NC 27599, USA

  • Venue:
  • Computer Communications
  • Year:
  • 2010

Quantified Score

Hi-index 0.26

Visualization

Abstract

In this paper, we propose a generalized framework for modeling the behavior of prominent congestion-control protocols. Specifically, we define a general class of loss-based congestion-control (LB-CC) mechanisms and demonstrate that many variants of TCP, including those being proposed for high-speed networks, belong to this class. Second, we develop a stochastic model to predict the transfer time for bulk transmissions by any protocol belonging to the LB-CC class-our model predicts both the mean as well as the variability in the transfer time. Our model is applicable to a wide set of transfer types and network capacities. We validate our model through extensive simulations under controlled settings, as well as with comprehensive HTTP workloads. We use our empirical analysis to also provide insights into several important issues, including: (i) identifying the settings under which previously-proposed TCP models are accurate, and (ii) identifying the conditions under which only steady-state analysis can be sufficient in modeling transfer performance. Our generalized framework provides a powerful tool that can be used in the design, analysis, and comparison of next-generation transport protocols. We demonstrate this benefit by comparing prominent TCP proposals for high-speed networks.