Multidimensional QOS Oriented Task Scheduling in Grid Environments

Abstract

High availability of resources is the key requirement in grids where computational and network resources are highly dynamic and unpredictable. QoS (Quality of Service) based resource management and scheduling is an important issue in such dynamic and heterogeneous grid environments. The computational resources in the grid vary from dedicated clusters to non-dedicated desktops which are connected over communication networks of diverse bandwidths. The task-resource mapping in such a non-deterministic computing environment leads to concerns over QoS constraints imposed by application tasks to identify suitable resources. In this paper, we consider bandwidth requirement as a QoS of a network and availability requirement as a QoS of a compute resource. An Availability aware QoS oriented Algorithm (AQuA) for task scheduling in grids has been proposed. The algorithm is evaluated and compared with QGMM (QoS Guided Min-Min) algorithm in the simulated grid environment. The experimental analysis and results show that the AQuA algorithm is able to utilize the highly available resources in grid and therefore increases the reliability to successfully execute applications without adversely affecting the makespan.