Scalable Management of Heterogeneous Cloud Resources Based on Evolution Strategies Algorithm

The emergence and rapid development of Cloud computing have intensively changed the Information Technology paradigm. Cloud computing found its wide application in science and engineering. Cloud-based systems provide new solutions for scientific experiments and access to data distributed across data centers. Optimization of Cloud resource management is becoming a significant area of research. This research focuses on scheduling tasks and processing scientific data in a heterogeneous Cloud environment where most jobs require a large number of resources and computing power. Our approach proposes a strategy for optimizing task scheduling across different virtual machines and data centers based on the metaheuristic Evolution Strategies algorithm. The Evolution Strategies algorithm was tested, which has not been used in this domain. As an essential property of the system, scalability has played an important role in selecting the algorithm. We created a model and added a Longest Job First broker policy. Compared to the standard Genetic Algorithm, our approach has shown improvements in measured metrics. After testing under different loads, the proposed strategy gave promising results and achieved a better makespan, larger average resource utilization, better throughput, less average execution time, a smaller degree of imbalance, and scalability.