challenges in Internet simulation & analysis


mathematical frameworks to find structure/patterns in traffic
a la scott's encouragement to `formalize some of what we (and providers) know' 
macroscopic as well as microscopic
theory of joing spatial/temporal locality
spectroscopy, tomography
ietf/ippm has been trying for a few years, but without dedicated funding

source modeling (for realistic inputs into simulations, models)
extract a set of source models from an aggregate trace 
feature extraction problem
10,000 gnutella port numbers are not 10,000 flows
ultimate goal: augment libraries of source level models w generation of own
calibrate models by evaluating their power for prediction

empirically validated simulation of a significant aspect of the Internet
already much work in large-scale simulations, but no recognized empirically validated simulation of any signficant piece of the Internet.
requires cooperation from providers and vendors to get default and configured parameters of OSes and algorithms.  NSF should shepard/foster this cooperation
(note: large scale means in size as well as # of protocols)