This code uses chirp packet trains (UDP) to estimate available
bandwidth (unused capacity) in the network.  The SENDER
(pathchirp_snd) and RECEIVER (pathchirp_rcv) modules run as daemons on
hosts at opposite edges of a path to be probed. Chirp packet trains
travel from the SENDER to the RECEIVER.

We initiate the pathchirp probing using a third module, pathchirp_run,
which can be run on any host (called the MASTER). This module also
collects the results of the experiment and stores them in a file
locally. Thus experiments can be initiated and the results analyzed
from a central location in the Internet which runs the pathchirp_run
module.

The pathchirp_snd and pathchirp_rcv modules can be started on
different machines around the Internet and will run indefinitely. They
need not be restarted for each experiment.


	         chirps
     SENDER ----------------> RECEIVER (pathchirp_rcv)
   (pathchirp_snd)	         |
			  	 |
      			      MASTER (pathchirp_run)
                 (starts experiment/collects results)




CORRESPONDENCE
==============

Please do send your comments about the performance of this available
bandwidth estimation tool, and any changes you feel would make the
tool more convenient or user friendly. Please let us know what
applications you need available bandwidth tools for so that we can meet
your needs best. The current tool is designed to be a low bit rate
tool for monitoring available bandwidth over long durations of time.

Contact: Vinay Ribeiro, vinay@rice.edu


COMPILE INSTRUCTIONS
====================
Run the following:
1) ./configure
2) make

The following will be generated:

Bin/<host_cpu>/pathchirp_rcv
Bin/<host_cpu>/pathchirp_snd
Bin/<host_cpu>/pathchirp_run

For example on certain Linux boxes <host_cpu>=i686.


RUNNING EXPERIMENTS
===================

1) SENDER (needs to be run just once)

Bin/<host_cpu>/pathchirp_snd

Options:
	 -U <sender port, default=8365>
	 -h 	Help: produces this output
	 -v version
	 -D print debug information

NOTE ON PORT NUMBER: In case the default port number is already used
at the sender, try out other port numbers larger than 1024 using the
-U option.  The receiver must then use the same port number with the -U
option.

2) RECEIVER (needs to be run just once)

Bin/<host_cpu>/pathchirp_rcv

Options:
	-h Help
	-v Version
	-D print debug information


3) MASTER (needs to be rerun for each experiment)

Bin/<host_cpu>/pathchirp_run -S <sender> -R <receiver> -t <expt. duration (secs)>

 ** In case of interrupt coalescence on Gigabit networks use the -J option:
Example:
Bin/<host_cpu>/pathchirp_rcv -S <sender> -R <receiver> -t <expt. duration (secs)> -J 6


All Options:

	 -n <number of estimates to smooth over, default=11>
 	 -d <decrease factor (>1), default=1.5>
 	 -b <busy period length (integer >2), default=5>
 	 -U <receiver port (chirp UDP), default=8365 (1024-65535)>
 	 -h Help: produces this output
	 -S 	 sender host name or IP address
	 -R 	 receiver host name or IP address
	 -J 	 number of packets per Jumbo packet,default=1. In case of packet coalescence use values larger than 1, e.g. -J 6
	 -l 	 lowest rate (Mbps) to probe at within chirp, default=10.000000Mbps. NOTE: by default pathchirp will find an appropriate probing range.
	 -u 	 highest rate (Mbps) to probe at within chirp, default=200.000000Mbps
	 -p 	 packet size <40-8200>,default=1000 bytes
	 -t 	 duration of experiment(sec), default=600 sec
	 -s 	 spread factor: ratio of consecutive inter-arrivals within a chirp, default=1.2
	 -a 	 average probing rate (Mbps), default=0.3Mbps
	 -v version


OUTPUT FILE
===========

At the MASTER, the file

	 <sender_name>_<receiver_name>_<time stamp>.instbw

will be generated in the directory from which the code is run.


FORMAT of the file:

 <time_stamp> <avail-bw estimate (Mbps)>