The Routing Over Large Clouds Mailing List Archive by date

[Scott Marcus <smarcus@bbnplanet.com>]

At 10:36 PM 5/29/96, Masataka Ohta wrote:

>> I think you folks just talked past each other.  Sam said, "over a simple
>> ATM star topology".  He was, if I understand him correctly, assuming that
>> the VCs converge at a single point in the "logical center" of the network.
>
>In WAN environment on Earth with N nodes and geographical size
>L, such topology costs O(L*N) for wiring, while mesh like one
>costs only O(L*sqrt(N)).
>
>That is, the architecture is impractical in large cloud WAN,
>where N is, by definition, large.
>
>Note that, in Sam's case, N is not so large.

Seems that we have another case of talking past each other.  I was talking
about a LOGICAL star, where the VCs converge on a central point; that says
nothing about the PHYSICAL network, i.e. how the wires run.

It's quite common for private Enterprise networks to be built as logical
stars of VCs, "hubbed" to one or two central routers over a carrier Frame
Relay infrastructure that is physcially a mesh topology.

There is no need whatsoever for excess length of the physical circuits.
The carrier shares the wires among multiple customers, and designs the
physical topology accordingly.




>> In the Frame Relay environment, this continues to be a very common and
>> practical way to construct internetworks, notwithstanding the hype you read
>> in the trade rags.
>
>I don't think Fram Relay is practical...

Paul Ferguson already responded to this much as I would have.  Large
numbers of these networks are deployed and running today.  They're doing
useful work, and they're operationally tractable.  No further proof of
their practicality is necessary.

And they are clearly within the charter of this group.

-- Scott Marcus