An Authorization Architecture Oriented to Engineering and Scientific Computation in Grid Environments

Large-scale scientific and engineering computation is normally accomplished through the interaction of collaborating groups and diverse heterogeneous resources. Grid computing is emerging as an applicable paradigm, whilst, there is a critical challenge of authorization in the grid infrastructure. This paper proposes a Parallelized Subtask-level Authorization Service architecture (PSAS) based on the least privilege principle, and presents a contextaware authorization approach and a flexible task management mechanism. The minimization of the privileges is conducted by decomposing the parallelizable task and re-allotting the privileges required for each subtask. The dynamic authorization is carried out by constructing a multi-value community policy and adaptively transiting the mapping. Besides applying a relevant management policy, a delegation mechanism collaboratively performs the authorization delegation for task management. In the enforcement mechanisms involved, the authors have extended the RSL specification and the proxy certificate, and have modified the Globus gatekeeper, jobmanager and the GASS library to allow authorization callouts. Therefore the authorization requirement of an application is effectively met in the presented architecture.