Diagnosing and Repairing Internet Performance Problems

David Choffnes earned his Ph.D. in Computer Science from Northwestern University in 2010 and is currently a postdoctoral research associate at the University of Washington. His research interests are primarily in the areas of distributed systems and networking, with a recent focus on mobile systems. He has coauthored three textbooks in computer science and programming, and has been awarded the CRA/NSF Computing Innovation Fellowship as well as the Outstanding Dissertation Award from Northwestern University

In this talk, I discuss how we can improve Internet reliability by enabling better tools for detecting, isolating, and repairing network problems as they occur. First, I discuss a system for crowdsourcing network monitoring to end. By leveraging the network view from applications running on a large number of hosts, we can efficiently detect network problems that impact end-to-end performance. Second, I describe a system for isolating the network responsible for a problem. To address this, we develop new tools that allow an ISP to identify the root cause even when portions of the Internet are unreachable. Third, I present an approach for automatically repairing isolated network problems. This allows an ISP to use existing routing protocols in novel ways to cause other networks to avoid problems, thus restoring normal connectivity.