Fault-Tolerant Scheduling Framework for MedioGRID System
Florin Pop, Dacian Tudor, Valentin Cristea, Vladimir Ioan Cretu · 2007
This paper presents a fault-tolerant scheduling framework that is mapped on the architecture of MedioGRID, a real-time satellite image processing system operating in a grid environment. Our study addresses the problem of fault-tolerant system for tasks execution. In MedioGRID, scheduling various computationally intensive and data intensive applications require information from satellite images. The proposed solution provides a fault-tolerant mechanism of mapping the image processing applications, on the available resources in MedioGRID clusters and uniform access. Various applications require simultaneous processing. Our approach demonstrates a very good behavior in the case of scheduling and executing groups of applications, while also achieving an optimized utilization of the resources in the system. We also survey the features of the Globus framework and the way these techniques can be applied on the MedioGRID architecture and highlight some solutions in order to enhance the fault tolerance at the MedioGRID application level, both simple and parallel.