Events & News
Ph.D. Dissertation Defense
| Category | Lecture |
| Date | Tuesday, May 20, 2014 |
| Time | 9:00 am |
| Concludes | 11:00 am |
| Location | Gould-Simpson 906 |
| Details | Review Committee: Drs. Beichuan Zhang, Chris Gniady, John Hartman & Richard Snodgrass |
| Speaker | Cheng Yi |
| Title | Ph.D. Candidate |
| Affiliation | Computer Science, University of Arizona |
Adaptive Forwarding in Named Data Networking
Named Data Networking (NDN) is a recently proposed new Internet architecture. By naming data instead of locations, it changes the very basic network service abstraction from “delivering packets to given destinations'' to “retrieving data of given names.'' This fundamental change creates an abundance of new opportunities as well as many intellectual challenges in application development, network routing and forwarding, communication security and privacy.
The focus of this dissertation work is a unique feature introduced by NDN: its adaptive forwarding plane. Communication in NDN is done by exchanges of Interest and Data packets. Consumers send Interest packets to request desired Data, routers forward them based on data names, and producers answer with Data packets, which take the same path of Interests but in reverse direction. During this process, routers maintain state information of pending Interests. This state information, coupled with the symmetric exchange of Interest and Data, enables NDN routers to detect loops, observe data retrieval performance, and explore multiple forwarding paths, all at the forwarding plane. Since NDN is still in its early stage, however, none of these powerful features has been systematically designed, evaluated, or explored.
In this dissertation, we present a concrete design of NDN's forwarding plane to make the network resilient and efficient. First, we design the basic adaptation mechanism and evaluate its effectiveness in circumventing prefix hijack attacks. Second, we propose a novel NACK mechanism for fast failure detection and evaluate its benefits in handling network failures. We also show that a resilient forwarding plane makes routing more stable and more scalable. Third, we design a congestion control mechanism, Dynamic Interest Limiting, to adapt traffic rate in a hop-by-hop and multipath fashion, which is effective even with a large number of flows in a large network topology.