Availability modeling of grid computing environments using SANs

In this paper, the availability of the Resource Management System (RMS) and computational resources distributed within grid computing environments is studied. Since the RMS acts as a heart of the grid environments, the unavailability of this system can render the entire environment to the inoperable phase. Furthermore, the unavailability of the grid resources may result in degradation of the performance of the grid. Therefore, considering the great importance of the availability issue in grid computing environments, the Stochastic Activity Networks (SANs) are exploited to model and evaluate the availability of grid environments. The proposed SAN models the failure of the resource management servers and grid resources, filling the queues of the RMS and resources and the arriving of the local and grid tasks to the resources, appropriately. Our proposed model not only can evaluate the availability of the RMS and grid resources, but also can consider the impact of the task scheduling algorithms on the grid availability. Simulations done on the hypothesis grid environments show the applicability of the model to the real grids.