Peering (and other petty bickering)

Michael Dillon michael at memra.com
Sat May 3 19:34:44 UTC 1997


On Sat, 3 May 1997, Sean Donelan wrote:

> >Look up ''peer'' in a dictionary, in this context it means something
> >like ''networks of equal size''.
> 
> I think you have cause and effect reversed.  A peer is someone you treat
> as an equal.  In fact, in the original meaning of "Peer," the Peers of
> the Realm (Barons, Lords, etc) were often extremly inequal in terms of
> property and power.  To end a the petty bickering that went on between
> various fiefdoms, the soverign conferred peerages.

A few moments spent reading http://www.baronage.co.uk/bphtm-02/moa-04.html
might be interesting if you have never investigated the history of
England, its laws, and its nobility.

There is also an intersting article which I read some twenty years ago
that bears upon this whole situation. I believe that it was published in
Scientific American and its title was something like "The Distribution of
Market Towns in Ancient Mesopotamia". It described the "network" of
trading patterns were able to learn of by excavating these towns.

In addition, it would be useful to study other networks found in both the
natural and the human world. For instance, the latticework skeletons of
tiny diatoms and the highway networks of North America.

Those who think they can understand networks by merely studying packet
switched data networks are wrong. They have no context and no perspective
for what they see before them.

If all the small elements in the structure of the Eiffel tower were
removed, the tower would collaps. Can you even imagine a road network that
does not have a full range of sizes in both the roads and the interchanges
and intersections?

This is very relevant to the discussion of peering because it relates to
how you design the network structures both physically and logically in
order to allow it to scale to ever greater sizes. Of course the Internet
is actually a complex interaction of several networks, i.e. the physical
topology is a network and the logical topology of BGP peering sessions is
another network layer seperate from, but intimately related to, the
physical topology. It is fairly obvious that you cannot build a highway
network in which every major highway in the USA meets in a single monster
interchange near Washington DC. And most people understand that it is not
scaleable to have all the traffic for every network provider flowing
through one NAP. IMHO, the same rule applies to the network of BGP peering
sessions as well. And it isn't yet clear to most people how we can scale
this aspect of the network. 

Now most people on this list seem to be jumping to the conclusion that
UUnet and the other majors are cutting back on peering for purely
commercial business reasons. But is this true? And how many of the
so-called business reasons actually are the direct result of technical and
engineering problems created by BGP itself?

Michael Dillon                   -               Internet & ISP Consulting
http://www.memra.com             -               E-mail: michael at memra.com

The bottom line is track record.  Not track tearing.  Not track derailing.
But pounding the damn dirt around the track with the rest of us worms.
       -- Randy Bush







More information about the NANOG mailing list