Peer-to-Peer Computing Transaction Middleware Services for Virtually Clustered GRID Networks using Information Resource Transaction Layer
 
Junseok Hwang and Praveen Aravamudham
 
CST 4-291, School of Information Studies, Syracuse University
Syracuse, NY, 13224, U.S.A
{jshwang, paravamu}@syr.edu
 
 
Abstract
 
 
Over the years, GRID computing has evolved into a de-facto standard for resource computations. Researchers belonging to a Virtual Organization (VO) are able to share data via a federated database system and aggregate their computational resources into a single pool via a resource broker. They are also able to run even the most demanding supercomputing tasks in parallel using the computing resources of the whole group. Members of this giant intranet are also able to buy CPU time from the others in the group - just like how power usage is metered. In theory, this is a perfect computing grid because it is based on flexible, secure, standards-based access to shared computing resources. However, in practice, successful clustering of VOs would require further research efforts on the issues like charging control, dynamic transaction accounting, metadata abstractions or reputation management among users who belong to multiple domains of Virtual Organizations. In this context, we introduce a core middleware architecture (Information Resource Transaction Layer (IRTL) capable of integrating p2p applications with resource and transaction specific services (GRID, Web Services, etc.) for providing the general capability of managing transactions with heterogeneous information resources; tools to transform the heterogeneous resource information into the metadata abstractions; Integrated APIs for peer-to-peer resource transactions; charging and accounting systems for peers and several other distinguishing features which lack with existing systems.