End-to-end availability between Internet hosts is between two and three orders of magnitude worse than other important engineered systems, including the US airline system, the 911 emergency response system, and the US public telephone system. This talk will explore why---access network failures, inter-provider and wide-area routing anomalies, domain name system faults, and server-side failures all have a role to play---and discuss how to improve the situation. I will describe two systems that can take advantage of the path redundancy available in the Internet and improve end-to-end availability by one or two orders of magnitude compared to the current state.

The key idea is to deploy an overlay network of nodes in different organizations and networks, have them carefully measure and monitor the status of available paths, and rely on them to cooperatively route packets by way of each other to bypass faults. This talk will describe two systems---RON (resilient overlay networks) and RAN (resilient access networks)---which use this core idea for different applications. RON is useful for collaborative multi-party applications, and RAN is useful for improving Web performance.

This talk presents an analysis of three years of data collected on the 31-site RON testbed, and studies the results from a six-site deployment of a RAN proxy that's been in active use for 9 months. The experimental evidence suggests that RON can reduce failures by a factor of six, and that with physical path redundancy, RAN eliminates almost all network-based failures.

Biography

David Andersen is a 5th year Ph.D. student in the Computer Science and Artificial Intelligence Laboratory at MIT. He received his M.S. from MIT in 2001, and B.S. degrees in Computer Science and Biology from the University of Utah. Before joining MIT, he was a co-founder and CTO of an Internet Service Provider in Salt Lake City. His research interests center on computer systems in the networked environment.