Files in this item



application/pdfUILU-ENG-03-2223_DC-211 assembled.pdf (858kB)Restricted to U of Illinois
(no description provided)PDF


Title:Primal-Dual Congestion Control Algorithms and E-Red AQM Scheme
Author(s):Liu, Shao
Abstract:In this thesis, we consider the primal-dual algorithm of network congestion control and E-RED AQM scheme as its implementation in the Internet. We first overview the recent work in network congestion control area, then introduce and analyze a decentralized network congestion control algorithm which has dynamic adaptations at both user ends and link ends, a so-called primal-dual algorithm. We show that this algorithm converges globally to the unique equilibrium, which solves the system problem in Kelly’s model in the absence of delay. We also give sufficient conditions for local stability of this algorithm in a general topology network with heterogeneous roundtrip delays. This local stability result suggests that a slow enough link adaptation stabilizes arbitrary fast source adaptation, and thus this local stability condition applies also for the dual algorithm. We next introduce a router algorithm called Exponential-RED (E-RED) as an implementation of the primal-dual algorithm in the Internet. E-RED is similar to RED except that it sets the marking probability as an exponential function of the instantaneous or average virtual queue length. Satisfying some conditions on the parameter choices, this E-RED algorithm can stabilize TCP-Reno and is shown to outperform RED by achieving higher bandwidth utilization and lower queueing delay at the same time. We finally provide some ns-2 simulation results to compare E-RED and RED and discuss about the influence to the performance of the parameter choices of E-RED.
Issue Date:2003-10
Publisher:Coordinated Science Laboratory, University of Illinois at Urbana-Champaign
Series/Report:Coordinated Science Laboratory Report no. UILU-ENG-03-2223, DC-211
Genre:Technical Report
Sponsor:National Science Foundation
Date Available in IDEALS:2018-04-03

This item appears in the following Collection(s)

Item Statistics