The new option --record allows you to record a session in two file, one
for each data direction. Then using the playback utility, you can
replay the information sent by the client (the .in file) to the server.
It will respect the data rate. The goal of this tool is to help
performance testing of a server. It is rather crude. Basically, instead
of pointing your client software to the host/port of the server, you
point it to a redir process in record mode. The redir process is started like
this

	redir --record /tmp/redirlog host local_port host_port

When a connection is made on the redir server on port local_port, it
connects to host on port host_port. It produces a file /tmp/redirlog.in.XXX
and /tmp/redirlog.out.XXX. XXX is the session ID (001,002,...). Each time
you connect to the redir server, it increments the session id. This way
you can record several session in parallel and each goes in its
own log file.

Then you use the playback utility to replay the session. This assumes
the server will behave exactly the same way. If a server behave differently
based on various outside event, the playback may not work. The solution
to this problem is either to modify your server so it fakes a bit
to produce a reproducible result. Or you use a smarter playback utility.

Anyway, to do the playback, you do

	playback host host_port /tmp/redirlog.in.001

To test the load on your server, you can start several playback process
at different interval.

This is a simple way to see how many clients a server can handle. Not
very precise, not that smart. Expect result with +-100% errors :-)