Development of innovative single and multi-objective metaheuristic methods for task scheduling in cloud systems

Unlike traditional computing units and data centers, cloud systems have a highly dynamic structure. The main factors that provide this dynamism are; customer types with many and different demands, the number of active virtual machines can change according to the instantly changing demand rate, and security requirements. Software and hardware services offered by cloud systems respond to this dynamism with the advantages provided by virtualization technologies. However, virtualization technologies are not sufficient to solve every problem that may arise. Inefficiently directing a large number of task requests from users to virtual resources results in high energy costs and resource waste for the provider, and expensive service purchases and dissatisfaction for the end user. A well thought out task scheduling strategy that ensures successful task distribution in cloud systems is of vital importance in this respect. Serious studies have been carried out in recent years for the task scheduling problem in cloud systems, which is quite complex, both in practice and in theory. This thesis study is studied for different scheduling strategies considering the dynamic nature of cloud systems. For this purpose, two innovative optimization methods are proposed for task scheduling in cloud systems. The first method is a single-objective optimization approach and aims to minimize the execution time of tasks in cloud systems. Unlike the existing methods in the literature, this method, which follows a rule based methodology, performs optimum scheduling by detecting the similarity between the task scheduling solutions in previously encountered scenarios and the current scenario. The success of this proposed method is presented comparatively with the solutions found by both classical scheduling methods and other metaheuristic methods. The results obtained in the experiments show that the proposed method is approximately 18 times faster than other methods in terms of decision-making process. The second method proposed in the thesis aims at simultaneous optimization of task execution time and energy consumption objectives. In this method, an approach based on parallel and hybrid operation of two different metaheuristic algorithms such as NSGA-2 and SPEA2 is used. The proposed method runs the two algorithms simultaneously and makes a selection according to their superiority over each other. The proposed method has obtained more successful results than the NSGA-2 and SPEA2 algorithms. Innovative and effective metaheuristic approaches on task scheduling in cloud systems have been presented in this thesis. This archive-based perspective is adaptable not only to cloud systems but also to similar dynamic systems. It is thought that the methods proposed in the thesis will contribute to the development of new solution ideas for researchers and that these approaches will be effective in different areas.