Machine Learning Technique for Energy, Performance and Cost-Effective Resource Management in Multi-Access Edge Computing

Abstract

Modern cloud interconnects on efficient resource allocation and provisioning to reduce their energy footprint. Resource management is complicated by factors such as data centre energy utilisation, virtual machine migration, operating expense, and overhead. Researchers have been using virtualized technologies and methods as optimal-Multi-objective particle swarm optimization, Dynamic Power Saving Resource Allocation (DPRA), Least Squares Regression, etc. to improve the management of their study. Accurately allocating resources to cloud users to meet their requests and offer QoS is a difficult task because of the preceding steps. Allocating cloud infrastructure's resources in the most efficient way possible benefits both users and service providers. The difficulties of resource management are tackled in this study by employing novel approaches, heuristics, authentication, and virtualization. In order to distribute workloads over several physical nodes, cloud computing relies on dynamic scheduling with load balancing. Using the help of host load prediction and a Markov chain model with Particle Swarm Optimization (PSO), VM resources are dynamically allocated to appropriate input requests. High quality of service (QoS) for cloud applications is achieved by SLA-based resource optimization with deadline, cost, storage, and bandwidth targets. Compliance with Service Level Agreements (SLAs), efficient use of resources, and low energy consumption are all achieved using a prioritisation technique based on SLAs. Scheduled users can receive resources in a predetermined order thanks to queuing. We developed the M/M/c/K queuing paradigm for numerous users per server to lessen the burden on data centres. Hardware resource models, such as CPU, I/O, and memory use, reveal VM resource allocation. Information gathering enhances resource utilisation and reduces energy consumption.