Cell Relay Archive

[fgoldstein@bbn.com (Fred R. Goldstein)]

In article <D78J6J.L9F@inter.NL.net> J.Lasschuit@aranea.nl (J.Lasschuit) writes:

>I agree that it depends on the application wheter a loss rate is acceptable 
>or not. But ATM was designed to work on high reliable media, with a very low 
>error rate. Otherwise, in those applications which are sensitive to errors, 
>one bit error leads to retransmitting a complete frame (a lot of cells). Of 
>course, one bit error in plain video or audio signals wouldn't do much harm, 
>but bear in mind that most of the video and audio signals are compressed 
>before sending. A bit error in compressed signals leads to unacceptable 
>noise in the signal (we want to have a clear noise free connection, don't we 
>?) And because those signals can't be retransmitted, bit errors are even 
>more a burden then within data. BTW: A cell loss ratio of percents is 
>certainly not what we expect from ATM.

Two problems still arise.  One, it's now acceptable to run ATM over 
practically any physical medium, from single-mode optics to unshielded 
twisted pair and almost to reindeer antlers.  (Isn't that the latest proposal 
from Lapland Telecom? :-) )  The original B-ISDN notion of glass only is 
obsolete.  And UTP in particular can easily be stretched to a high BER.

Far more important, though, is congestion-related loss.  CBR won't throw away 
anything unless a) it's broken, or b) you exceed your SCR.  VBR won't throw 
away _much_ unless a) it's broken, or b) you exceed your SCR or PCR, or c) you 
get unlucky and some bursts converge in the network and overflow a buffer 
("funneling" loss).  ABR won't throw away anything unless a) it's broken, or 
b) you violate its policy, or c) they take some brain-damaged rate-based hack 
and call it ABR.  UBR won't throw away anything unless a) you use UBR, in 
which case all bets are off and a .7%, 7%, or 75% loss rate are all totally 
reasonable to expect, depending upon instantaneous load of the network.

Among all of those, I'd say it's unreasonable to expect ATM networks to have 
extremely low loss rates at all times.
___
Fred R. Goldstein   k1io    fgoldstein@bbn.com
Bolt Beranek & Newman Inc., Cambridge MA  USA   +1 617 873 3850
Opinions are mine alone; sharing requires permission.