Exchanges that matter...

Fri Dec 6 20:32:15 UTC 1996

salo at msc.edu (Tim Salo) wrote:

>> FDDI is not a WAN technology.  Losing 20% of bandwidth of a 10 ft
>> piece of fiber is one thing.  Losing 20% of bandwidth of a
>> $3M/yr circuit is quite different.

>I don't understand.  Are you trying to say that price/performance
>analysis is appropriate for LAN technologies, but not for WAN technologies?

Quite opposite.  Price of "transmission" in LANs is negligible as
compared to the price of interface cards and switches.  So a good
price/performance of the overall system can be achieved by simplifying
those switches or interfaces, particularly by using less efficient
methods of transmission.

The price of transmission facility in WANs easily exceed price of switching
equipment by an order of magnitude or more.  Therefore it is very silly
to sacrifice transmission capacity for the sake of easier switching in
WANs.

>Now, I could understand if you said, "I analyzed ATM WAN services for
>one configuration once and concluded that other technologies provided
>better cost/performance in all possible configurations," or even
>"I don't think ATM products and standards are mature enough for my
>application."

I don't think it will work as anything but sync muxes with silly framing
in the real life WANs.  I published my reasons for thinking so.

>However, you seem to be saying that wide-area technologies
>should be assessed based on total cost and overhead, rather than on
>cost/performance...

No, i simply left out a simple conjecture.

--vadim