Abstract

The Internet plays a crucial role in interconnecting billions of devices, users, and applications. To serve its purpose, the Internet infrastructure should accommodate the ever-more-demanding performance requirements of new applications (streaming video, VoIP, algotrading, and beyond). Yet, despite extensive efforts by both researchers and practitioners, the Internet infrastructure has remained fairly stagnant for decades and, consequently, suffers from both bad performance and alarming security vulnerabilities.

I will discuss two promising approaches for bootstrapping innovation in the Internet: (1) exploiting network programmability at Internet eXchange Points (IXPs), the crossroads of Internet traffic, to provide new network functionalities, and (2) enhancing the routing expressiveness of today’s legacy networks without modifying the network infrastructure. As an example of (1), I will present SIXPACK , a theory-informed approach for route-computation at IXPs that both provides richer expressiveness in route selection and is provably privacy-preserving. As an example of (2), I will present COYOTE , a legacy-compatible traffic engineering mechanism that is robust to traffic variations.