The Internet is not only radically changing the way we communicate, educate, and conduct research and business, it is transforming the very fabric of our society. The Internet offers both tremendous opportunities and challenging problems. One important problem is network mapping. Network maintenance, provisioning, design, quality of service, and security all benefit from accurate mappings of network connectivity and performance as functions of space and time. To date, a wide variety of Internet maps have been produced using existing networking tools. Most of these standard mapping techniques, however, provide only a partial picture of the Internet. The decentralized nature of the Internet makes quantitative assessment of network performance very difficult. Due to many factors, including the expense of special-purpose measurement equipment, the burden of transmitting measurements from internal network nodes to a central processing center, and confidentiality and proprietary issues, mapping methodologies cannot depend on the cooperation of individual computers and routers to freely transmit vital network performance statistics such as traffic rates, delays, and dropped packet rates. This leads to the following intriguing question: Can accurate maps of network performance be derived from limited numbers of measurements at a limited set of measurement sites? This talk will describe our recent efforts to shed light this question.
1:00pm
McMurtry Auditorium, Duncan Hall 1055
For additional information, contact:
Last Updated: September 26, 2002