Thursday, February 27th, 2003
Room 104, Gates Computer Science Building

Some Not-So-Pretty Admissions About Dealing with Internet Measurements

Vern Paxson
ICSI Center for Internet Research (ICIR) at Berkeley

About the talk:

If we look "under the hood" of any large Internet measurement study, the view isn't pretty: we find everything from inconsistencies to systematic errors to lost analysis chains to intractable data navigation to absent calibration to misconceptions about what's being measured to irreproducible results. The net result is that as an empirical science, Internet measurement is sorely lacking. We discuss three fundamental ways to strengthen the field's discipline: working towards a culture of calibration, creating an emphasis on reproducible results, and fostering the availability of high-quality datasets within the community.

About the speaker:

Vern Paxson is a senior scientist with the ICSI Center for Internet Research (ICIR) at the International Computer Science Institute, a non-profit research institute in Berkeley, Calif. He is also a staff computer scientist at the Lawrence Berkeley National Laboratory (LBNL). His research focusses on network intrusion detection, Internet attacks, and Internet measurement. He serves on the editorial board of IEEE/ACM Transactions on Networking, and as chair of the Internet Research Task Force, as well as program co-chair of ACM SIGCOMM 2002, program chair of USENIX Security 2003, and on the steering committee of the Internet Measurement Workshop. He received his Ph.D. from the University of California, Berkeley.