ITR: Unifying Software Frameworks for Distributed Computing

Emory University, Atlanta GA

Investigators

Abstract

An architectural framework for distributed computing is proposed, with a view to unifying multiple resource sharing models. Farm computing, private virtual machines, computational grids, and peer to peer technologies are examples of di.erent paradigms that have evolved to support di.erent application domains. However, despite their common underlying theme of sharing resources across computer networks, systems implementing these technologies tend to preclude interoperability and, in some cases, require substantial middleware to orchestrate resource aggregation. This project adopts a provider-centric view of resource sharing, emphasizes lightweight software infrastructures that maintain minimal state, and facilitates straightforward interfacing to current and emerging distributed computing standards. Through a reconfigurable, component-based approach, the proposed framework will attempt to provide support for multiple distributed computing paradigms in a unified setting. The provider perspective adopted in this project envisions resource owners hosting a software backplane onto which owners, clients, or third-party resellers may load components or component suites that deliver value added services. Complete control over usage, security, and protection of resources is retained by the owner. Components may implement primitive functions such as message transport or mathematical transforms, or may be aggregated to emulate concurrent computing environments, or may ofer portal interfaces to complex backend application services. Providers of primitive resources or components may publish their availability using either standard mechanisms, framework-supplied schemes, or private directories; clients discover, lookup, and use these services subject to usage and security constraints. In the conduct of this project, fundamental issues concerning commonalities and dualities in distributed systems, secure resource sharing, and software reconfiguration will be explored, and will likely contribute new and useful findings to the field. Owing to its strong pragmatic bias, the project is also likely to have broad impact by suggesting schemes to realize effective distributed computing systems, and by evolving prototype paradigmatic software.