Recent advances in processing (in particular, network processors) and storage provide the technological foundation for a new generation of continuous-monitoring measurement systems that can continuously and selectively capture, process, and store packet/packet-header information for high-bandwidth links for periods of time ranging from hours to days. Use of these measurement systems, both individually and collectively will in turn enable a rich new set of real-time, adaptive network-management and traffic-profiling capabilities, and further enhance our fundamental understanding of network phenomena. In this proposal, we describe fundamental, cross-disciplinary research aimed at the design, development, and application-use of a next generation of distributed, high-performance passive network measurement infrastructure. Our proposal brings together researchers with the needed expertise in hardware and architecture, operating systems, databases, and various aspects of networking (measurement, protocol design and analysis, routing, security, and application-level overlays).
Our project, which we refer to as Hyperion, divides broadly into four efforts (i) Our first focus is on the design, development, and prototype implementation of a new adaptable, resource-aware, network measurement node architecture based on the use of emerging network processor (NP) chips. (ii) While a single measurement node necessarily provides a myopic network view, a system of Hyperion nodes will enable a much richer set of network-management and traffic-profiling capabilities. With multiple Hyperion nodes comes the attendant concerns of node synchronization, node placement, node coordination, and distributed data compression. (iii) We also explore a number of application-level uses of Hyperion nodes, including fault-detection, overlay support, and traffic characterization. This application-use will help define and refine Hyperion APIs, overall functionality, and internal structure and operation. (iv) An important part of our proposed research will be to work with network providers (of our departmental networks, a tier-II regional network, and a tier-I backbone network) to understand network-management needs, and demonstrate Hyperion capabilities. We will also distribute a number of Hyperion nodes to external researchers, and make data we collect available to others.
