Cell Relay Archive

[albert.e.manfredi@boeing.com]

In article <3607A4C9.705047FD@mars.dgrc.crc.ca>,

> Dewayne, Bert,
>
> I have read your discussion with interest. While not able to follow all
> its details, the bottom line, as I understand it, is that when the
> bandwidth requirement exceeds that provided by some physical layer
> standard, e.g., 3 x STS-1 (OC-3, ~155 Mb/s), but is less (often much
> less) than the bandwidth provided by another with a much higher
> capacity, e.g., 12 x STS-1 (OC-12, ~620 Mb/s), what options are there?
>
> Would "Inverse Multiplexing for ATM" (IMA) be an option?
> Of course, at this time, IMA is under study by the ATM Forum.

IMA might be an answer, but I think that the issue in that discussion went
beyond just what to do when you exceed the STS-3c limit. I mean, that problem
exists in all networks. What do you do when you exceed the 10 Mb/s capacity
of Ethernet? Do you just go up to a 16 Mb/s Token Ring or a 25 Mb/s ATM net,
or do you jump to 100 Mb/s Ethernet? Big jumps are the norm, not the
exception.

I think the real issue is that a non-concatenated STS-12 is not the answer. A
non-concatenated STS-12 is still a 622 Mb/s link. Non-concatenated simply
means that the SONET frames inside this STS-12 superframe are to be separated
out into their individual STS-1 and/or STS-3c pipes. Demultiplexed, SONET
style. But the bit rate arriving over this OC-12 link is still 622 Mb/s,
although perhaps much of that bandwidth is not being used. Wasted, in other
words, if you aren't using ATM to carry the data over SONET. That's precisely
why ATM was invented!

On the other hand, if you used ATM switches at the merge points of the SONET
infrastructure, now that last OC-12 link to the "central ATM switch" would be
STS-12c, so that the data from each of the branches merging into that last
link could make use of the excess bandwidth whenever possible. The real value
of this comes when you add more branches to this tree structure, and can get
by without increasing that last STS-3c link. Possible if you use ATM, but not
if you use basic SONET pipes only.

Alternatively, perhaps if you use ATM switches at the merge points instead of
simple SONET nodes, that last link can remain STS-3c at least for awhile. It
all depends on what type of traffic you're handling. If it's mostly data, that
might be perfectly adequate. Heck, we've been merging three Ethernets onto one
256 Kb/s FR WAN connection for several years now, and things work pretty well.
ATM can do the same, and do it better. SONET by itself cannot.

The idea of wanting SONET node functionality incorporated inside ATM switches
is viable, I guess, but it doesn't seem to me to be a great step forward. I
guess what we really need is for the carriers to provide ATM at merge points.

Bert
manfredi@arl.bna.boeing.com

-----== Posted via Deja News, The Leader in Internet Discussion ==-----
http://www.dejanews.com/rg_mkgrp.xp   Create Your Own Free Member Forum