Fault Tolerant Job Scheduling And Resource Management Scheme For Economic Based Computational Grid

Grid Technology offer's the capability of aggregating and sharing resources at a very large scale that are geographically distributed across different locations and organization boundaries. Grid resource management is the process of identifying requirement for grid applications, matching resources to these applications, allocating those resources, and monitoring Grid resources over time in order to run grid applications as efficiently.In large scale grid, lack of centralized controlled environment, predominant execution of long jobs, highly dynamic resource availability, diverse geographical distribution of resources from different administrative domains, and heterogeneous nature of grid resources, exponentially increases probability of failures and their impact on grid performance degradation, when compared with the traditional parallel systems. Thus, the incorporation of fault tolerance related features in job scheduling policy mechanism is not an additional optional feature, but a necessity. In this thesis we have investigated the issues related to resource allocation, fault-tolerance and application scheduling in the context of computational grid economy, arises due to the autonomous nature of grid environment. Our proposed approach provides a novel solution for fault-tolerant resource allocation and task scheduling algorithm that executes user's job in case of resource failure and satisfied user quality of service (QoS) requirements.We have thoroughly reviewed the different resource management systems (RMS) for different existing grids presented in literature and discussed motivations for the use of computational economy as a metaphor for the management of resources and application scheduling in Grid computing environments. We have investigated the implementation of computational economies driven brokering system to be applied to peer-to-peer computing applications and grid systems infrastructure for encouraging people to share files, contents, or music in larger scale by providing them economic incentive.We have proposed an economic model for resource selection for economy based grid and a fault-tolerant scheduling algorithm that selects appropriate resources to execute user's jobs to satisfy their Quality of Service (QoS) requirements i.e. within allocated budget and deadline while considering resource failure during job execution. Our strategy used tuple-space based approach to provide fault-tolerance using transaction and check-pointing techniques. If resource fault occurred i.e. a resource is not able to execute the tasks, our strategy relocate the task to another resource and starts its execution from the point where it interrupted using check-pointing technique hence reducing the overall makespan of user's job. We have evaluated the performance of our scheduling strategy against Work Queue (WQ), Work Queue with replication (WQR) and Time Optimization based scheduling strategies through extensive simulation. We used the GridSim toolkit-4.0 to simulate a Grid environment. The detailed performance evaluation of our fault-tolerant algorithm is carried out through a series of simulations by varying a number of performance metrics like users, deadline, budget, fault index, optimization strategies and simulating geographically distributed Grid resources.From simulation results we have showed that our scheduling strategy performs better than WQ, WQR and Time Optimization scheduling strategies in term of number of tasks completed within deadline and budget parameters and overall reducing the makespane of tasks. From the measured results we conclude that, our strategy provides a suitable solution to fault-tolerance scheduling in economy based grid environment because proposed strategy has shown visible improvement in satisfying the user QoS requirement, thus proposed strategy has made economy based grid environment, more reliable and consistent, even in the presence of resource faults. Hence proposed strategies helps in sustaining user's faith on grid by satisfying his QoS parameters i-e deadline & budget, by enabling grid to deliver reliable and consistent performance in the presence of faults.