Scalable Management of Heterogeneous Cloud Resources Based on Evolution Strategies Algorithm

Abstract

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.

A Comprehensive Survey on Sustainable Resource Management in Cloud Computing Environments

“…the optimization algorithm operates by iteratively refining the mapping solutions to achieve the best allocation of VMs on servers. Therefore, the optimization algorithm fine-tunes the mapping of VMs to Heuristics and machine learning approaches that improved energy efficiency servers, striving for the most efficient allocation [205]- [208]. The value of the objective function, guided by the minimization or maximization of specified objectives, encapsulates the efficiency and suitability of the resource allocation [209]- [212].…”

Efficient algorithm for replanning web service composition

“…Not only that, but in DAG applications, neither [2] nor [3] have they addressed any strategy for decreasing energy consumption during a specific operation. Clouds are considered heterogeneous if platforms offer sufficient storage space for effective computation of workloads and support distributed computing, such is the case with cloud-computing [4], [5].…”