In order to reach their large audiences, today's Internet publishers
primarily use content distribution networks (CDNs) to deliver content.
Yet the architectures of the prevalent commercial systems are tightly
bound to centralized control, static deployments, and trusted
infrastructure, thus inherently limiting their scope to ensure cost
recovery.
This talk describes a number of techniques (and the resulting systems)
that realize highly-scalable cooperative content distribution. I show
how to build a fully self-organizing CDN that efficiently delivers
content using unreliable nodes, describe how to transparently dispatch
clients to nearby servers, and touch on how to securely leverage even
untrusted resources.
These ideas have been implemented, deployed, and tested in production
systems currently serving several terabytes of data to more than a
million people every day. The view of the Internet gained from deploying
these systems answered long-standing questions about network
configuration, while providing important insights that helped evolve our
systems' designs.
Bio: Michael J. Freedman is a doctoral student at NYU, visiting Stanford
since 2005. He received his M.Eng. and S.B. degrees from MIT. His
research interests span distributed systems, security, networking, and
cryptography, with a focus on both principled designs and real
deployments.