Learn

challenge: measure and analyze data on Internet infrastructure
that yield insights into workload, performance, growth,
or stability of the overall system

Introduction

Sustained, rapid growth, increased competition, and proliferation of new applications have combined to change the workload profile of the Internet in recent years. Sheer traffic volume and link capacity have rendered monitoring and analysis challenging. Following its involvement with the NSFNET backbone project and transition away from that backbone's central role in U.S. infrastructure, the NSF has remained dedicated to supporting research on the increasingly complex, and now largely commercial, Internet system. The National Science Foundation has sponsored workshops on Internet Statistics Measurement and Analysis (ISMA) since 1996, and also sponsored (National Laboratory for Applied Network Research) as part of the national R&E vBNS project, and CAIDA (Cooperative Association for Internet Data Analysis), in their network measurement, analysis, and visualization research.

CAIDA offers this set of web pages to attempt to convey results of these publicly funded efforts to the community.

real-time flow characterization
- CoralReef monitor for OC2 and OC12 ATM links (till ported to other interfaces/L2-protocols)
- latest (Inet'98) paper on Coral statistics (using data from internet MCI backbone)
- coral-dev mailing list (subscribe using coral-dev-request@caida.org)
- Traffic traces from FIX.west (few-minute samples, headers with aliresses masked) - No longer available
- 5 minutes of destination IP addresses only (No longer available) from FIX.west around noon sat 2/22/97, for use for router designers that want samples of address locality
NASA-Ames collaborative activities (how we get much of the raw data)
CAIDA's Internet measurement tool taxonomy

really important note, especially for the litigious

Internet traffic data in these pages derives from samples we take from the most intuitively relevant places to which we have access. We can't offer any guarantee or even probability that they are representative of traffic in other locations or Internet traffic in general (whatever that is).

Even worse than the fact that the data is a sample whose representativeness is suspect, we typically process the raw data with programs written in perl or C, almost always by a human, and thus vulnerable to errors. So, given limited available data on the nature of the Internet beast, and the elusive search for invariants in its workload profile, embarking on high-investment activities like building routers or really expensive systems grounded in these data would be imprudent at best.
(Too bad we don't have anything better, huh.)

Data brought to you via the support of the National Science Foundation, and cooperation and support of NASA-Ames, MCI, Digital Equipment Corporation, ANS, and may others involved in https://www.caida.org.

Everything you've learned in school as "obvious"
becomes less and less obvious
as you begin to study the universe.
For example, there are no solids in the universe.
There's not even a suggestion of a solid.
There are no absolute continuums.
There are no surfaces. There are no straight lines.
-- R. Buckminster Fuller

Questions/Comments Info@CAIDA.org