Wide-area Dissemination under Strict Reliability, Timeliness, and Cost Constraints

A National Science Foundation (NSF) grant (Award Number 1535887, September 2015 - August 2018) to Johns Hopkins University. A component of the NSF Algorithms in the Field (AitF) program.
Principal Investigator: Michael Dinitz. Co-Principal Investigator: Yair Amir.

Overview

This project aims to enable new Internet services with demanding timeliness and reliability requirements by developing new theory and a practical architecture for resilient routing. Emerging interactive applications, such as remote manipulation, have such strict timeliness requirements that reliability cannot be achieved solely through buffering and recovery protocols, as packets may arrive too late to be useful. While redundant sending along a single path and network coding can improve reliability, the combination of correlated loss on the Internet and the strict timeliness constraints of target applications dramatically reduces the effectiveness of such approaches.

This project seeks to provide timeliness and reliability by sending each packet over a subgraph of the network, rather than along a single best path, and to select subgraphs that are cost-effective, which is essential to enable these services to become widely available in practice. There is currently no way to optimally trade-off reliability and cost, so we are developing new algorithms to solve the problem at several levels of abstraction, and are implementing and testing the performance of these algorithms on a global wide-area network to evaluate the practicality of the abstractions and the effectiveness of the solutions.

Contributors

Related Publications