Vern Paxson

Dr. Vern Paxson's PhD dissertation in the mid-1990s laid the groundwork for the exciting resurgence in research in Internet measurement during the last ten years. Through a series of highly influential papers, Dr. Paxson not only provided a wealth of useful information about the nature of the deployed Internet, but also raised the practice of network measurement to an entirely new level by combining the extensive collection of data from many locations with sophisticated statistical techniques. His findings about the Internet, and most importantly the mismatches between reality and the assumptions commonly made in analytical and simulation models, led the networking community to place much greater emphasis on the collection and analysis of measurement data as a major part of the research process.

Dr. Paxson's SIGCOMM'94 paper on "Wide-area traffic: The failure of Poisson modeling" (co-authored with Sally Floyd) showed definitively that the (then common) practice of modeling packet and connection arrivals as a Poisson process significantly underestimates the burstiness of real Internet traffic. His SIGCOMM'96 paper on "End-to-end routing behavior in the Internet" analyzed more than 40,000 measurements between 37 Internet sites to study pathological routing behavior, such as instability, loops, and outages, and demonstrated how to draw meaningful statistical inferences from indirect observations of the Internet's routing system. This paper recently won SIGCOMM's first "test of time" award, given to the most influential networking paper written ten years ago. Dr. Paxson's SIGCOMM'97 paper on "End-to-end internet packet dynamics" analyzed 20,000 bulk file transfers between 35 Internet sites to study the frequency and effects of path asymmetry, out-of-order packet delivery, packet corruption, and packet-loss patterns, including the important finding that loss events have much more complex statistical structure than previously thought.

Dr. Paxson's PhD thesis -- a tome at over 400 pages -- is one of the only dissertations that was (and still is) widely read by other researchers in the networking community. His early measurement papers are still widely cited, and his style of research adopted as a gold standard for how to measure a complex, heterogeneous network like the Internet and make statistically sound statements about its properties and their implications. The vibrancy of the Internet Measurement Conference (which he co-created in 2001) and the prevalence of measurement papers at other networking conferences are a testament to the influence of his initial research in this area.