Internet Address Space Clustering for Intelligent Route Control

An intelligent route controller optimizes traffic from a subset of the
Internet (IP) address space to a set of (usually) non-overlapping regions 
called clusters.  Each cluster has a set of associated measurement points 
(scanpoints) which may be, but are not necessarily, identified by 
addresses within the cluster.  To make routing decisions from the optimized 
network to a cluster, the route controller performs a series of 
measurements to the scanpoints over multiple Internet connections.

This talk will focus upon a set of techniques used to cluster the
Internet Address space.  The approach combines structural, topological, 
and temporal techniques for clustering and differs from previous approaches 
by supporting variable granularity, i.e., the opportunity to trade 
complexity for accuracy.