Order Tray | Contact Us | Home | SIG Lists

[Ham-80211] Aerocomm Modems: PPP Experiment

Guillaume Filion gfk at logidac.com
Tue Oct 25 01:09:16 UTC 2005


Hi all,

I'm interested in setting up a 900 MHz link using the Aerocomm modems, 
you may remember that I did a small experiment a couple weeks ago to 
estimate the throughput that I would get with real-time compression:
=====
That gives between 36% and 38% compression. For a 78 Kbps link that
would raise it to about 121 Kbps for plain text.
=====

I did another experiment with PPP this time, I set up a NSLU2 single 
board computer (slug) and linked it to a PC running linux using a cross 
over serial cable. I then setup a PPP connection and did some 
benchmarks, with impressing (impossible?) results.

You can read the whole article online at:
http://guillaume.filion.org/blog/archives/2005/10/connexlink_900_1.php
and here's the part that puzzles me:
=====
[...]I then followed the instructions of the Serial-Laplink-HOWTO.

I was surprised to see how easy it was, I just copied and pasted the 
code example, changed ppp_laplink_server to ali's IP address and 
ppp_laplink_client to the IP address that I wanted to give to the slug 
via the PPP link (different than the one it got from the ethernet). 
Voilà! It worked on the first try!
-----
root at ali:/etc/ppp# /usr/sbin/pppd /dev/ttyS1 nodetach
Using interface ppp0
Connect: ppp0 <--> /dev/ttyS1
PAP peer authentication succeeded for gfk
Deflate (15) compression enabled
found interface eth0 for proxy arp
local IP address 192.168.0.2
remote IP address 192.168.0.43
-----
root at slug:/etc/ppp# /usr/sbin/pppd /dev/ttyUSB0 nodetach
Using interface ppp0
Connect: ppp0 <--> /dev/ttyUSB0
Remote message: Login ok
PAP authentication succeeded
Deflate (15) compression enabled
local IP address 192.168.0.43
remote IP address 192.168.0.2
-----

Benchmarks

I tested the reliability of the link with a 15 minutes ping flood:
-----
gfk at ali:~$ sudo ping -f 192.168.0.43
PING 192.168.0.43 (192.168.0.43) 56(84) bytes of data.
.
--- 192.168.0.43 ping statistics ---
74602 packets transmitted, 74601 received, 0% packet loss, time 997772ms
rtt min/avg/max/mdev = 15.099/19.923/39.925/0.239 ms, pipe 3, ipg/ewma 
13.374/19.929 ms
-----

Then I measured the bandwidth with iperf. Note that those speeds are all 
half-duplex.
-----
gfk at ali:~$ iperf -c 192.168.0.43
------------------------------------------------------------
Client connecting to 192.168.0.43, TCP port 5001
TCP window size: 16.0 KByte (default)
------------------------------------------------------------
[ 5] local 192.168.0.2 port 43315 connected with 192.168.0.43 port 5001
[ 5] 0.0-10.0 sec 3.65 MBytes 3.05 Mbits/sec

I was a bit stunned by these results -- getting 3 Mbps on a 0.12 Mbps 
link is a little too good to be true -- but I realised that iperf must 
have been using really simple data for its test. And everyone should 
know that simple data compresses really well. When using "real world" 
data or even compressed data, I get more realistic figures:
-----
gfk at ali:~$ iperf -c 192.168.0.43 -F maildir.tar
------------------------------------------------------------
Client connecting to 192.168.0.43, TCP port 5001
TCP window size: 16.0 KByte (default)
------------------------------------------------------------
[ 6] local 192.168.0.2 port 43312 connected with 192.168.0.43 port 5001
[ 6] 0.0-10.4 sec 400 KBytes 314 Kbits/sec
gfk at ali:~$ iperf -c 192.168.0.43 -F maildir-best.tar.gz
------------------------------------------------------------
Client connecting to 192.168.0.43, TCP port 5001
TCP window size: 16.0 KByte (default)
------------------------------------------------------------
[ 6] local 192.168.0.2 port 43310 connected with 192.168.0.43 port 5001
[ 6] 0.0-11.5 sec 136 KBytes 96.9 Kbits/sec
-----

These numbers are still very impressing, in my planning I estimated that 
I would get about 37% of speed gain, but these "real world" results give 
around 224% of speed gain. If anyone has an explanation, I'd be happy to 
hear it.
=====

So, if anyone has an explanation, I'd be happy to hear it! :-)

Thanks and 73,
GFK's
-- 
Guillaume Filion, ing. jr
Logidac Tech., Beaumont, Québec, Canada - http://logidac.com/
PGP Key and more: http://guillaume.filion.org/




More information about the ham-80211 mailing list