Experiments with Distributed Component Computing Across Grid Boundaries

Abstract—There are situations when a user has simultaneous access to computing resources in many virtual organizations built on different Grid infrastructures, as well as in local clusters or single machines outside Grids. Such a user may be interested in running a large-scale application on all available resources simultaneously, avoiding huge administrative overhead. We argue that in such situations the component model which can support parallel execution of distributed components may be very convenient, especially for scientific applications which go beyond pre-installed modules. We show how to exploit the dynamic runtime nature of CCA model, and define mechanisms for creating components with parametrized numbers of ports. As a user-centric method of resource aggregation across Grid boundaries, we create a virtual pool of H2O kernels, spawned on the resources, using such protocols as SSH or batch schedulin