부산대학교 도서관

With increasing demands for data storage, processing and advanced computing, cloud computing has become almost indispensable in different domains of engineering and technology. However, cloud-based platforms are resource constrained and need to distribute the tasks in such a way that the workload is balanced. This becomes more streamlined if the estimate for cloud workload is available beforehand. Hence task scheduling and workload forecasting are non-trivial tasks which need to be executed meticulously. In this paper, a heuristic approach for task scheduling has been proposed along with a machine learning based approach for workload forecasting. The performance of the task scheduling approach has been evaluated in view of the response time and CPU utilization. The performance of the workload forecasting approach has been evaluated based on the regression, mean absolute percentage error, mean square error and number of iterations for training.