Why Understanding Anything About the Internet is Painfully Hard

Monday, November 22, 1999 - 17:30
TH 331
Vern Paxson Lawrence Berkeley National Labs

The Internet poses immense challenges when attempting to measure, model, analyze, or extend it. The underlying problems are that the network is tremendously diverse, such that the notion of "typical" is rarely anything but misleading; it grows explosively, and has been doing so for years; and the traditional network modeling framework, based on Poisson assumptions of independence or at most weak traffic correlations, lies in ruins, destroyed by overwhelming measurement evidence. We sketch the fundamental causes of these difficulties and some of the strategies researchers have devised for coping with them.