Timely-NS3

This is a Timely implementation based on TCP congestion control module in ns3
https://github.com/ychen884/Timely-NS3

Build

  1. NS3 version 3.33
    Please follow the instructions below to build ns3-3.33

https://www.nsnam.org/releases/ns-3-33/documentation/

https://www.nsnam.org/wiki/Installation

After building successfully, you should be able to see two file paths:

../ns-3.33/src/internet/model/..

../ns-3.33/scratch/..

Replace all the files in these two folders with the files in this github repo under model and scratch folder

For someone not sure how to configure the building script:

Edit the file …/ns-3.3/src/internet/wscript

  • line 323: headers.source: add model/tcp_cc_timely.h in bracket
  • line 108: obj.source: add model/tcp_cc_timely.cc in bracket

Then you can directly build with ./waf inside ns-3.3, and you should be ready to run.

If this does not work or you are looking for a simpler approach, I would suggest downloading my virtual machine image to have a try.

  1. You can get download ubuntu vm from here [https://drive.google.com/drive/folders/1pagHFs4arInrFA-4-tnZQaMQuKCltMrZ?usp=sharing]
    install the vm image with vmware/vbox/…

cd ~/ns-allinone-3.33/ns-3.3

./waf –run “scratch/test –congestion=TCPCCTIMELY –incast=10” for simple demonstration

  1. Scenerio: a simple incast topology with one sink and N source, with limited egress bandwidth.
    Please check the comment in scratch/test.cc for more detailed topology description.

Run

  1. Run basic program

Example:
running 10 nodes sending into one sink node with a egress bandwidth 25Mbps

/waf –run “scratch/test –congestion=TCPCCTIMELY –incast=10”

The command below runs the same parameter as the default ones, but explicitly set the options. Check this to learn how to use these cmd options.

./waf –run “scratch/test –congestion=TCPCCTIMELY –incast=10 –Alpha=0.6 –AI=1.0 –MD=0.05 –initial_rate=5 –Hth=4500 –Lth=500 –HAI=5”

CMD argument options with default value inside bracket[]:

--congestion:	Protocol to use: TCPCCTIMELY, TcpVegas, TcpDctcp [TCPCCTIMELY]; Notice all other supported protocol can also work, but requiring one to modify the source code becuase I add call back to trace RTTs
--h:         	Showing man [false]; --h=true shows the man page
--bandwitdh: 	egress bandwidth [25Mbps]
--incast:    	incast num [15]
--Alpha:     	Alpha [0.6]; EWMA weight, the larger weight, we put less value to historical data. Range: [0,1]
--AI:        	AI [1]; Additive increasing parameter
--MD:        	MD [0.05]; Multiplicative decreasing parameter
--initial_rate:  initial_rate [5]: initial sending rate
--Hth:       	Hth [4500]: Higher threshold
--Lth:       	Lth [500]: Lower threshold
--HAI:       	HAI counter [5]: HAI counter, number of completion events with negative gradient that triggers HAI mode
  1. Export txt stat file and show the graphs:

./waf –run “scratch/test –congestion=TCPCCTIMELY –incast=10” >> output_25_10.txt (you can set any output file, and any running cmd options if you want)

Make sure you have python 3.8 with matplotlib installed

Then run the python script main.py, make sure the output file is in the same directory

I have provided the main.py in /Draw and my txt output file to generate my graphs in the /Graph (LF be replaced by CRLF becuase of github)

GitHub

View Github