A Decision-Making Based Resource Provisioning Model in Cloud Computing Platform

Introduction: Cloud computing enables the efficient and scalable processing of large-scale data but faces challenges in aligning resource provisioning with dynamic workloads. Traditional strategies often cause overprovisioning or underprovisioning, which can impact performance and energy efficiency. This study proposes a clustering-based approach for organizing tasks by specific features, coupled with time-series forecasting to predict resource demands per task cluster. The aim is to achieve efficient, energy-aware Virtual Machine (VM) provisioning for fluctuating workloads. Methods: The Decision-Making-based Cloud Service Provisioning (DM-CSP) model enhances scheduling and allocation through three stages: task clustering, workload analysis, and timeseries- optimized provisioning. Tasks are grouped by arrival time, deadline, and computation time. Within each cluster, the ARIMA model forecasts workload intensity, enabling proactive VM allocation. A novel task-to-VM mapping strategy minimizes energy use while meeting deadlines. Results: Tests on real-time datasets show that DM-CSP improves execution time, reduces energy consumption, and increases deadline compliance. Performance indicators such as MAPE, RMSE, throughput, and correlation coefficients confirm the robustness. Discussion: DM-CSP effectively addresses workload volatility by integrating clustering and predictive analysis for proactive, energy-efficient provisioning. Its performance and security features make it scalable and reliable for multi-cloud environments. Conclusion: DM-CSP outperforms conventional methods by improving task completion, optimizing resource utilization, and enhancing Quality of Service (QoS). Trust-based mechanisms further ensure secure, reliable interactions, positioning DM-CSP as a dynamic, scalable architecture for high-performance cloud environments.