Re: 28.8 Upgrade

Ko Stoel (ko@komac.knoware.nl)
Sat, 3 Jun 1995 22:10:50 +0200


>you wrote
>>At 14.4, SLIP connections show screens, but very slowly, infact NASA
>>doesnt work for me. With 28.8 SLIP, will I be able to view the NASA screen?
>>Or broadcast at 28.8?
>
>During my research I investigated a variety of dial up telephone line
>connections using both 14.4 and 28.8 SLIP and PPP connections. My research
>showed that PPP was more reliable, and 28.8 provides passible performance,
>when used in conjunction with Talk Window. A lot also has to do with the
>quality of line that your service provider gives you access to the net
>with, and this quality usually depends on how much you want to pay.
>
>df

To add to this;

It can be quite confusing as where in the chain there's a bottleneck:

framegrab => compress => slip/ppp => route|.. => [Reflector] => \
display <= decompress <= slip/ppp <= route|.. <= <= /
possible other participants <= /

possible limiting factors;

1. compress/uncompress your machine can't cope , has to limited cpu power
2. slip/ppp and route are unpredictable; what will happen from router to router
3. Reflector mediates the average througput from all receiving participants
to the sending CU-SeeMe front-end; So if you are the only modem user, for
instance, and the others have T1 going at top speed THEN you get UDP
bombarded or at least the routers their buffers will fill-up even if you
disconnect you get all the extra buffered UDP stuff as a after-meal. Vice
versa lot's of modem users slow down T1 to T1 (the faster users). Also even
if you haven't any window open you will get 1kbps per participant
control-data from the Reflector

To test for the best modem PPP/SLIP connection you can use ping with udp as
protocole and with a representant content of the packets; this prevents you
from variing CU-SeeMe data and makes comparing things easier. You can ping
to your dial-up gateway or any router underway or even the reflector yes
even the other CU-SeeMe front-end. Thus step by step you will have insight
where the first bottleneck is.
Ping times the roundtrip time and this constitutes delay formed by
bandwidth fysical delay and software delay. Delay is the time the first
datum gets back, bandwidth decides on how long it takes to complete the
test-packet. Whatever theorethical underlayments here; do the test with
ample packets of considerable size look at the average roundtrip time and
repeat the test with another modem/protocole or ping-victim.

PPP has more overhead than SLIP is more reliable but with good POTS
phone-lines there's no need for it
Modem compression (MNP) could even be a disadvantage for allready
compressed data.
Also the MTU (slip parameter) value plays a role
Of course 28k8 is better than 14k4 .
Don't be fooled by IP providers stating 'We have 28k8 dial-up', If there's
some obstruction in the chain, no 28k8 modem will help you.

What should be a representent ping packet I don't know yet.

And then there are the CU-SeeMe parameters ;
Change Tolerance etc. Ah..... sound.

It all looks like pre-Tford cars lots of handles knobs and bulky repair kits :)
weren't they fun.

--
Ko Stoel
Amersfoort
Holland