Grid systems deployment & management using Rocks

Wide-area grid deployments are becoming a standard for shared cyberinfrastructure within scientific domain communities. These systems enable resource sharing, data management and publication, collaboration, and shared development of community resources. This work describes the systems management solution developed for one such grid deployment, the GEON Grid (GEOsciences Network), a domain-specific grid of clusters for geological research. GEON provides a standardized base software stack across all sites to ensure interoperability while providing structures that allow local customization. This situation gives rise to a set of requirements that are difficult to satisfy with existing tools. Cluster management software is available that allows administrators to specify and install a common software stack on all nodes of a single cluster and enable centralized control and diagnostics of its components with minimal effort. While grid deployments have similar management requirements to computational clusters, they have faced a lack of available tools to address their needs. We describe extensions to the Rocks cluster distribution to satisfy several key goals of the GEON Grid, and show how these wide-area cluster integration extensions satisfy the most important of these goals.