Effective priority-based resource allocation for proactive auto-scaling framework in workload prediction using hybrid tree-enhanced vector machine model

Abstract

Abstract Cloud computing (CC) is a paradigm offering on-demand access to software, platforms, and infrastructure as a service (IaaS). IaaS enables cloud providers to supply resources like Virtual Machines (VMs) and storage, making resource management critical for minimizing costs and maximizing profits. To optimize resource allocation, this paper proposes a priority-based auto-scaling framework using machine learning (ML) for workload prediction. The framework has four phases: monitoring, analysis, planning, and execution. In the monitoring phase, data is collected to assess cloud resources. The workload for each VM is classified using a hybrid tree-enhanced vector machine to determine resource utilization. The planning phase utilizes a Hybrid Vultures and Waterwheel Plant Optimization (H-VWPO) model to decide on auto-scaling policies. Finally, in the execution phase, cloud customers’ requests are distributed to VMs based on priority, optimizing VM utilization and response times. Implemented on a PYTHON platform, the proposed model achieves superior performance, with accuracy improving from 98.68% for 15 VMs to 99.12% for 50 VMs.

Keywords

• Resource allocation
• Machine learning
• Proactive auto-scaling
• Priority-based
• Workload prediction