Simulations for the "Promoting the Use" paper

The page contains the tcl scripts used for running the simulations in the following paper. The scripts can also be gotten from the tar file. The README file describes the procedures for running these simulations.

[FFT98] Sally Floyd, Kevin Fall, and Kinh Tieu, Promoting the Use of End-to-End Congestion Control in the Internet , February 1998.


These scripts have been verified to run with ns version 2.1b2 and with ns-2.1b6, and should run with any releases of ns-2 later than 2.1b2.
Problems of unfairness:

The simulations for Figures 2 and 3, "Simulations showing extreme unfairness with three TCP flows and one UDP flow, and FIFO scheduling" and "Simulations with three TCP flows and one UDP flow, with WRR scheduling. There is no unfairness", can be run in ns-2 with Fairness.com and supporting scripts Fairnessall.v2.tcl, Fairnessall1.v2.tcl, Collapse.v2.tcl, and Setred.v2.tcl. (The scripts on the web site were only recently updated to include Figure 3 as well as Figure 2.)


The danger of congestion collapse:

The simulations for Figures 4-7 can be run in ns-2 with Collapse.com and supporting scripts Fairnessall.v2.tcl, Fairnessall1.v2.tcl, Collapse.v2.tcl, and Setred.v2.tcl.

The simulations for Figure 8 were made with an earlier version of the simulator.


Characterizing TCP-friendly flows

The simulations for Figure 10, "TCP-friendly bandwidth for a 60-ms roundtrip time and 1460-byte packets" can be run with Dropallold.tcl, with supporting scripts Dropband.tcl and Setred.v2.tcl. (These can be plotted using S with "csh Drop1.com", or this can be plotted on a log-log scale, with the UMass version of the TCP-friendly equation added, using S with "csh Drop2.com".

A modified version of this chart on a log-log scale, showing SACK (line 1), Tahoe (line 2), and Reno (line 3) TCP, along with both the simple and UMass versions of the TCP-friendly equation, can be run with Dropall.tcl, with supporting scripts Dropband.tcl and Setred.v2.tcl. (These can be plotted on a log-log scale using S with "csh Drop3.com".

The figure drop4.ps compares the simple and UMass versions of the TCP-friendly equation with a modified version of the UMass equation that was hoped to be a more accurate model for SACK TCP in the moderate-packet-drop regime. This figure can be created with "csh Drop4.com".

Note that for all of these figures, the x-axis shows the fraction of packet drops - multiple packet drops in a single window of data have not been counted as a single congestion event, as one would like to more closely match the TCP-friendly equation.


Return to [ Sally Floyd].

floyd@icir.org
Last modified: January 2001. Links updated on October 2008.