Policy and Mechanism in the Future Internet

Michael Walfish: Assistant Professor, Department of Computer Science, University of Texas, Austin

Video of lecture

Abstract

Michael Walfish

Policy is now crucially important for network design: there are many stakeholders, each with requirements that a network should support. Among many examples, senders have an interest in the paths that their packets take, providers have analogous interests based on business relationships, and receivers want to shut off traffic from flooding senders. There is of course a bewildering bevy of architectural proposals already, but they each target a subset of policy concerns, and they cannot generally be implemented together. Our research has asked: can we empower all stakeholders and uphold all reasonable policy considerations with a single mechanism?

Our tentative answer is yes. We have designed a network called ICING. ICING places routing and transit policies under the control of end-hosts and provider-operated general-purpose machines. ICING’s data plane exposes only four primitives. ICING does not require centralized trust. Perhaps most importantly, and unlike past policy proposals or the status quo, ICING strictly enforces network policy. For example, ICING binds the data plane to the dictates of the control plane. An implementation of ICING in hardware shows that ICING is feasible, though more expensive than IP.

Joint work with David Mazieres, Jad Naous, Antonio Nicolosi, and Arun Seehra.

Bio

Michael Walfish is an assistant professor of computer science at The University of Texas at Austin. He received his B.A. from Harvard in 1998, worked in industry for four years, received his PhD from M.I.T. in 2008, and did postdoctoral stints at Stanford and University College London in 2008 before joining UT, where he has been for a year and a half. His research interests are in networked systems. He is very much looking forward to visiting UC San Diego.