Current Internet network controls are realized by an ad hoc combination of protocols, automated configurations, and manual configurations. This ad hoc environment makes it difficult to provide concrete behavioral assurances. Moreover, the entire system might exhibit harmful emergent behaviors that are difficult to anticipate, test for, debug, or correct.
The architecture for the future Internet should thus eliminate the current ad hoc practices and start from a clean slate. This project is developing a new architecture in which each network control function is realized as an application that runs on top of an operating platform. The operating platform provides crucial abstractions, interfaces, and services to allow applications to effect changes in the underlying network nodes, to enable the composition and concurrent executions of multiple applications, and to provide protection mechanisms to guard against potentially harmful actions of applications. Intellectual challenges such as scheduling, synchronization, inter-application communication, resource multiplexing, and operational invariant protection for the new breed of network control applications are being investigated.
Broader Impacts: This project directly addresses a key challenge for future networks, namely network manageability. Problems in network manageability affect the health of the networks we depend on, and they affect the manner in it is possible for innovation and economic growth to occur with these networks. Our new approach to manageability, the operating platform architecture, can potentially stimulate a market for a variety of 3rd-party network control applications. Finally, this project provides particularly exciting opportunities for the training of undergraduate and graduate students because they develop applications in a fundamental way, but addressing high-impact current issues. We expect to diffuse the findings from this project into cutting-edge courses on network management.
