[ppml] Address Space versus Routing Slots
tli at tropos.com
Tue May 9 03:52:15 EDT 2006
> Aggregation is NOT the solution. The need for aggregation is proof
> that we have failed to develop a scalable routing architecture. Why
> are so many people determined to preserve this failure? It makes no
> sense to me. Why not look for a scalable routing solution that can
> be depolyed without requiring aggregation?
A scalable routing architecture requires that the information content
that is being exchanged scale at a rate that is substantially less (in
order theoretic terms) than the number of nodes and sites that are being
attached to the network. Abstraction of nodes into groups of nodes is
the only possible method of significantly reducing the required amount
of information. In routing terms, the abstraction of addressing (more
specifically, locator) information is aggregation. Any other
abstraction of locator information would also be aggregation. This is
true for both geographic abstraction, topological abstraction, or
"everyone who's surname starts with R" abstraction. All abstractions
are likely to require exceptions. The number of these exceptions
necessary to achieve adequate connectivity defines the efficiency of the
abstraction scheme. Some abstractions are more efficient than others.
Abstractions are frequently hierarchical (i.e., computer science tree
structured), as a multi-level abstraction allows for gains in efficiency
at higher levels and recapture of losses due to exceptions.
A routing architecture is 'scalable' if the cost of running the
architecture is 'reasonable'. Precise definitions of these terms is a
judgement call. However, we have another technology known as
'bridging', where no abstraction occurs and the amount of control
information scales linearly with the number of nodes in the network.
This is currently thought to not be scalable to the size of the
More information about the ARIN-PPML