ettcp is an updated version of ttcp

Features added:

Timed test mode, allowing easy measurement of
network throughput where the speed is not known
in advance (e.g test for 10 seconds, rather than
for 1000 buffers).

Callback mode, where ettcp makes a connection
to the remote node, then expects to _receive_ data.
This removes the need to run another copy of ttcp
at the remote end. We can use a dumb xinetd server
to spew data at whoever connects.

Many compiler warnings removed.

Two shell scripts: speedfrom and speedto which
make it simple to run a 10 second speed test.

Simple xinetd compatible programs "suck" and 
"blow" which respectively receive data and transmit
data on stdin and stdout.

xinetd.d files for suck and blow

There are obvious potential denaial of service
issues with these--- consider using xinetd
to restrict access to these ports.

Enhanced result printing code: it is now possible
to make ettcp quit from printing chatty output, it
only prints the network speed measured. Raw format
where only the bits/s number is printed, without
"Mb/s" etc. Useful where a program is parsing the
output.

This code has only been tested on 2.4.x versions
of Intel Linux. But there's no obvious reason
why it should't work for other OSes and architectures.


Text below is the README file from ttcp
------------------------------------------

TTCP is a benchmarking tool for determining TCP and UDP performance 
between 2 systems.

The program was created at the US Army Ballistics Research Lab (BRL)
and is in the public domain. Feel free to distribute this program
but please do leave the credit notices in the source and man page intact.

Contents of this directory:

ttcp.c		Source that runs on IRIX 3.3.x and 4.0.x systems
		and BSD-based systems.  This version also uses getopt(3) 
		and has 2 new options: -f and -T.

ttcp.c-brl	Original source from BRL.

ttcp.1		Manual page (describes ttcp.c options, which are a 
		superset of the other version).


How to get TCP performance numbers:

	receiver				sender

host1%  ttcp -r -s			host2% ttcp -t -s host1

-n and -l options change the number and size of the buffers.