An Efficient Job Scheduling Technique in Trusted Clusters for Load Balancing

Although there has been tremendous increase in PC power and most of it is not fully harnessed, yet certain computation intensive application tend to migrate their process with the aim of reducing the response time. Cluster computing is the area that aims just at this. Clustering provides means to improve availability of services, sharing computational workload and performing computation intensive application. However, these benefits can only be achieved if the computing power of cluster is used efficiently and allocated fairly among all the available nodes. As is the case with our desktops, it is usually seen that clusters also suffer from underutilization. A number of approaches proposed in the past share only idle CPU cycles and not use the resources of systems when the machine has its own local processes to execute. We propose a priority-based scheduling approach for run queue and multilevel feedback queue scheduling approach for migrated tasks that doesn't degrade the performance of local jobs too. Simulation and experimental results have been able to show that priority-based run queue management and multilevel feedback queue scheduling for migrated tasks can increase overall throughput by about 28-33 percent.