GridWeaver: automatic, adaptive, large-scale configuration for Grid computing

The Gridweaver project reviewed the management tools and techniques currently available for Large Scale Configuration. We substantiated that these will not be able to sustain the demands of the Grid services currently in production, unless there is a paradigm shift towards high-level, declarative descriptions of fabric configurations. Our prototype configuration architecture offered this power and allowed our research Grid service, GPrint, to demonstrate ease of maintenance, adaptability to the demands of the fabric and fault tolerance. This paper introduces system configuration as a major issue to the e-Science community and presents the findings of our research. Interviews of representative large-scale installation managers and our projected case studies for future use of the Grid are distilled to form the requirements of Grid fabrics. We present our prototype architecture, combining the consortium’s background approaches to system configuration. We then discuss our GPrint Grid service, which demonstrates the successful investigation of some of the key configuration deficiencies identified by the consortium.