ISMA 2002 workshop:
Multiresolution and correlation analysis of global Internet measurements

Final Report
http://www.caida.org/workshops/isma/0210/
http://www.lc.leidenuniv.nl/lorentz_center/2002/20021007/info.php3?wsid=59

After last year's (2001) workshop on `Multiresolution Analysis of Global Internet Measurements', coordinated by Ogielski and Cybenko with great success, CAIDA pursued a similar collaboration with the Lorentz Center for its annual ISMA workshop. The Lorentz Center demonstrated such success in 2001 as a faithful vehicle for building sustainable channels of intellectual exchange for the Internet research community, that CAIDA wanted to explore the possibility of further establishing such a channel of collaboration among researchers in the American and European, especially Dutch, communities.

The Lorentz Center environment allowed CAIDA to hold a workshop it could not hold at its home institution in the United States; the unique environment of the Lorentz Center facilitated more of a `retreat' or even `sabbatical' milieu rather than the conference/symposium motif so typical of this and many other fields.

The content of this year's workshop focused on Internet inter-domain routing and topology analysis. Talks ranged from discussions of the performance of the root Domain Name System to fine scale quantification of BGP behaviors. Representatives from industry (Interroute, Global Crossing, Akamai, Agilent, Sprint, IBM) provided valuable calibration to `real world' Internet behavior and practice for academic and laboratory researchers from both the U.S. and Netherlands. The overall theme and discussions centered around analysis of current inter-domain routing system behavior, in hopes of grounding a superior design of the next generation's Internet routing system. Prepared talks and follow-up discussions also covered current and proposed community data collection efforts (a key such current effort is based at www.ripe.net in Amsterdam), as well as verification/calibration across both the data sets and analyses of that data over the last two years.

The commerical participants provided insight into how providers overprovision their backbone topologies, and how external routing policy as well as performance measurements affect the engineering of both backbones as well as CDNs (content distribution networks). The mathematicians provided reminders, in the way of inquiry, of the gap that persistently remains between where we are now and a formal model of the real Internet.

One interesting discussion touched on the recent cybersecurity-focused federal government organizations in the U.S., interested in garnering a `synoptic' view of the Internet. While everyone admits that we do not have a precise definition of what that means, all agree that the first step is to attempt to characterize `normal' Internet behavior, from which we can begin to taxonomize various behaviors we might consider `abnormal', in terms of stability or service quality.

Topics introduced with preliminary investigation but requiring future attention include:

1. correlations between:
   1. DNS and BGP (e.g., ccTLD request behavior with routing changes)
   2. forward [traceroute] path and reverse [BGP-announced]
   3. forward and reverse AS paths
   4. routing changes and TCP/UDP performance
   5. large-scale Internet worm activity and effects on BGP behavior
2. predictive topology-based indicators of performance
3. more sophisticated analysis of large scale topologies

Other topics discussed included clarification of prevailing mis-assumptions about routing table growth and churn, and strategies for testing BGP/router performance in a test lab (with suitable forwarding load on the test router).

CAIDA may collaborate on another workshop in Leiden in 2003, in particular to help foster collaboration for a recently funded community project: `Correlating Heterogeneous Measurement Data to Achieve System-Level Analysis of Internet Traffic Trends' seeking substantial input (participants) from the mathematics and statistics community. Anyone interested in participating in such a workshop please send mail to info@caida.org with your specific interests and current/imminent work in the area.

---