The Computing Innovation Fellows Project

Posted on: Thursday, May 26th, 2011
Broad Research Area: Networks / Operating Systems, Theory / Algorithms

Research Interests:

My research spans networked systems and theory. Ongoing projects include:

- Low-latency networking: Even simple application-level tasks in today’s networks, like retrieving a small web object, take many times longer than the underlying network latency. How can we design the Internet’s protocols to approach speed-of-light responsiveness? We are exploring novel low-latency designs in naming and transport protocols, and in scalable routing on flat names. There is interesting work to be done ranging from theoretical analysis to fundamental protocol design to applications including Content-Centric Networking, Internet architecture, and ad hoc networks.

- Network debugging: We are building a tool, Anteater, that examines data plane state of networks in order to find bugs in a flexible, unified manner. A paper on Anteater will appear in SIGCOMM 2011, and there are very exciting opportunities for further work in this area.

- Data center network topologies: With researchers at UIUC and UC Berkeley, we are studying how to build high-bandwidth data center networks that allow incremental expandability, and are simultaneously more efficient than traditional topologies like fat trees. A short paper on our initial design, Jellyfish, will appear in HotCloud 2011.

- Future Internet Architectures: The Internet has become the basis of most of humanity’s remote communication. Its rapid growth and the fact that we use it in so many unexpected ways attest to the groundbreaking flexibility of its design. Yet because of this growth, the Internet’s architecture has encountered a range of limitations. What would an Internet look like that provides the dependability, security, manageability, and efficiency to enable innovative applications for decades to come? With collaborators at Berkeley, Georgia Tech, MIT, Princeton, and Stanford, we are designing and developing an architecture that answers that question.

- Flexible Networks With Source Control: Giving users some control over packet routing is a promising technique to dramatically improve reliability, efficiency, and flexibility of networks, but it is also a radical shift from today’s Internet. How can a routing and forwarding architecture flexibly support source control, while respecting the interests of network owners? How can users quickly pick reliable, efficient routes that are appropriate for the application, given limited information about the dynamic state of the network? What is the effect of source control on denial of service vulnerability? Answering these questions critically requires tools ranging from planetary-scale systems building to game-theoretic analysis.

- In addition to the above topics, my past work deals with randomized algorithms, peer-to-peer systems, incentive compatibility, distributed systems, and more; I maintain my interest in these and other areas of networked systems and am open to collaborations on various topics.

A particular feature of my research is to integrate practical network systems design with theory. This approach leads to systems built on firm foundations with guarantees on their behavior, as well as interesting and difficult theoretical problems that can expose fundamental principles. As such, I’m actively seeking fellows who are system builders, theoreticians, or interested in bridging the two areas.

Contact Information:

pbg@illinois.edu