Resource selection using regression techniques

Grid scheduling is the process of making scheduling decisions involving resources over many administrative domains. Resource selection is termed as resource discovery, assignment of application tasks to resources, and data staging. The Grid scheduler does not control the set of jobs submitted to it, or even know about jobs being sent to resources so decisions that trade-off one job's access for another's cannot be made in a global sense. Makespan represents lapsed time from the first task's beginning to the end of the last scheduled task. If the Makespan is small, the utilization of the machines is high. Design space for Grid Scheduler's is usually rich. The scheduler must choose carefully between different implementations of user authentication, allocation, and reservation. The objective in this study is to minimize overall job completion time or the application Makespan, the latter often being the performance feature in resource allocation study. Makespan minimization arranges tasks to level differences between each work phase's completion time. Techniques combined into heuristic approaches or in upper level multi-objective methodologies (i.e., meta-heuristics), are the lone methods to schedule when there is a high problem dimension and complexity. As optimization techniques, metaheuristic are stochastic algorithms trying to solve many hard optimization problems that are effective than traditional methods. So, recent scheduling problem research focused on such techniques. This aim of this research is to examine meta-heuristic approaches for scheduling issues as the latter is a NP complete problem. In this paper Performance evaluation of resource selection using SVR with proposed GA optimization and proposed hybrid GA optimization is investigated. The proposed SVR with RBF kernel and hybrid GA optimization decreases the mean absolute error by 9.63% than SVR with RBF kernel and GA optimization and by 22.15% than SVR with RBF kernel.