A model of grid service capacity

Computational grids (CGs) are large scale networks of geographicallydistributed aggregates of resource clusters that may be contributed bydistinct organizations for the provision of computing services such asmodel simulation, compute cycle and data mining. Traditionally, thedecision-making strategies underlying the grid management mechanismsrely on the physical view of the grid resource model. This entails theneed for complex multi-dimensional search strategies and a considerablelevel of resource state information exchange between the grid managementdomains. In this paper we argue that with the adoption of serviceoriented grid architectures, a logical service-oriented view of theresource model provides a more appropriate level of abstraction toexpress the grid capacity to handle incoming service requests. In thisrespect, we propose a quantification model of the aggregated servicecapacity of the hosting environment that is updated based on themonitored state of the various environmental resources required by thehosted services. A comparative experimental validation of the modelshows its performance towards enabling an adequate exploitation ofprovisioned services.