Internet Systems Measurement and Analysis Projects at UW

Web, caching systems, CDN, and P2P studies

[ Publications | People | Local web ]

Understanding the behavior of currently deployed systems is important for understanding how future systems should be designed. An ongoing research thrust here at the University of Washington is to measure and analyze interesting aspects of the Internet and World-Wide-Web. For example, we have studied proxy caching, peer-to-peer file-sharing systems, and malicious threats, such as spyware.

Web document sharing and cacheability: we have been studying web document sharing and cacheability of web documents. The key to our studies has been an "organizational-based" analysis. Using UW as a model of a large collection of independent, diverse organizations (we've identified about 200 independent organizations with the university), we have studied web document sharing both among clients within individual organizations, and among clients in different organizations. Using this data and an analytical model, we have also evaluated in some detail the potentials of inter-proxy cooperation (cooperative caching), and have examined the temporal properties of streaming media workloads flowing into the University of Washington.

Peer-to-peer document sharing systems: We have performed several detailed measurement studies of popular P2P file-sharing systems, including Napster, Gnutella, and Kazaa. For example, we characterized the population of end-user hosts that participate in Gnutell and Napster. This characterization includes the bottleneck bandwidths between these hosts and the Internet at large, IP-level latencies to send packets to these hosts, how often hosts connect and disconnect from the system, how many files hosts share and download, the degree of cooperation between the hosts, and correlations between these characteristics. Our measurements show that there is significant heterogeneity and lack of cooperation across peers participating in these systems, both of which should (but have not yet) drastically influence the design of P2P sharing systems. We also performed some of the first quantitative studies of Kazaa, demonstrating important properties of P2P workloads. We also created a model that allows synthetic generation of parameterized P2P workloads.

Spyware: We performed the first quantitative study of spyware by looking at how widespread several common adware programs are on hosts at the University of Washington. In a more recent study, we have crawled the Web looking for spyware piggy-backed on executables and also drive-by-downloads, in which scripted Web pages install spyware through the browser.




Current students

  • Charlie Reis
  • Roxana Geambasu
  • Cherie Cheung
  • Alexander Moshchuk
  • Tanya Bragin
  • Richard Dunn

Former students

  • Stefan Saroiu
  • Krishna Gummadi
  • Geoff Voelker
  • Alec Wolman
  • Molly Brown
  • Tashana Landray
  • Felix Livni
  • Denise Pinnel