Comparative Analysis of Simulated Cloud Task Scheduling

Cloud systems are autonomous systems, that are managed on the base of complex set of criteria. In clouds systems, where large number of tasks runs on numerous machines, optimized task scheduling causes significant reducing of computing time. The process of scheduling tasks in a cloud computing system consists of two steps. The first step is to allocate computing tasks to virtual machines and the second step is to create virtual machines on physical machines. In two steps scheduling takes place, so we talk about combinations of scheduling models. We introduce a complex analysis of the behavior of a simulated cloud system to find appropriate combinations of scheduling models for scaled and dimensional sets of computational tasks. The tasks have different requirements on the system, such as those that are focused on the number of computational cores. We use the minimum completion time of the last task as the main quality criterion. Based on experiments, 9000 simulations, and subsequent comparative analyses, we find that the scheduling of tasks to virtual machines has a significant impact on makespan. We have shown that the optimized task scheduling (with minimum makespan values), reduces the 6 original classes of combined models to 2 classes. These are on time sharing and space sharing models for scheduling tasks on virtual machines.