PARTAGE: An Open Peer-to-Peer Infrastructure for Cycle-Sharing

Purdue University, West Lafayette IN

Investigators

Abstract

Hu/Vitek abstract Large-scale compute-cycle sharing for complex and large scientific applications is a long standing challenge in distributed computing. Many cycle-sharing systems have been developed and have delivered orders of magnitude improvement in available resources. Nevertheless they are limited in their performance or capability. Furthermore, other forms of resource sharing and coordination are needed to achieve truly distributed access to scientific infrastructure. The goal of this project is contribute to the national science and technology infrastructure by developing enabling technologies for virtual laboratories open to researchers from all disciplines. This infrastructure will leverage resources in peer-to-peer computing and programming languages to deliver an alternative architectural model for resource sharing over wide area networks. The self-organizing aspects of peer-to-peer networks promise to eliminate much of the difficulty required to configure and maintain a large-scale globally distributed computing and control system, and decentralization removes the potential bottleneck and single point of failure of centralized service nodes. This project's contributions will be to provide abstractions for Peer-to-peer Computing based on extensions to the object-oriented programming model and to develop a Peer-to-peer Infrastructure which extends the current peer-to-peer networks developed for routing and data-centric applications. This project will work collaborativey with the DARPA Program Composition for Embedded Systems (PCES) initiative to apply the proposed resource-balancing techniques developed in this project to sensor networks and distributed real-time systems of national importance. The educational impact of the proposed research includes integration of advanced distributed systems research into both undergraduate and graduate curricula, and outreach activities that involve high schools students into active participation in Internet-scale computing.