NeTS-FIND: A Swarming Architecture for Internet Data Transfer

University Of Massachusetts Amherst, Amherst MA

Investigators

Abstract

Data transfer is a defining task of a network. This project, uswarm, proposes a novel data-transfer architecture inspired by recent, tremendously successful swarming systems such as BitTorrent. The project pursues the following, rather exploratory, question: can swarms form the basis of a universal data-transfer architecture for a future Internet and if so, what is an appropriate architecture? Natural systems with swarm-like properties are known to be extremely robust both for an individual and for the system as a whole. The technical motivation is to make Internet data transfer: 1) scalable, 2) fault-tolerant, and 3) incentive-compatible, a trio of properties difficult to achieve using existing solutions. The research underlying uswarm will: 1) develop an in-network BitTorrent-like substrate deliver practically all data, not just large files, 2) exploit in-network storage and locality of data accesses, 3) analyze availability and human wait-time improvements of one large integrated swarm as opposed to BitTorrent-like isolated swarms, 4) analyze new traffic engineering knobs and routing choice naturally enabled by uswarm, 5) develop novel incentive strategies for integrated swarms accounting for network-level incentive issues, 6) analyze the interaction of selfish behavior with routing and congestion control using game-theoretic techniques. Broader Impact: If successful, this project will fundamentally improve Internet data transfer by enabling: 1) critical information dissemination resistant to denial-of-information attacks, 2) delay-tolerant network applications, and 3) novel social networking applications. The proposed work includes undergraduate as well as international teaching and outreach activities integrated with the proposed research activities.