The era of cloud computing has created the need for intelligent placement of workloads in cloud computing environments due to variations in workload demand, cloud pricing, latency, and availability of resources among providers and industries. Hence, the paper introduces the AI-Driven Intelligent Decision Support Framework for Multi-Cloud Resource Management Using Workload Prediction and Cost-Performance Optimization. The framework incorporates workload traces, machine metadata, cloud pricing, and latency-availability information to predict CPU workload demands and facilitate cloud-provider selection. Five regression models were put into consideration, namely: XGBoost, Random Forest, Decision Tree, Gradient Boosting, and Extra Trees. The models were tested using the 5-fold cross-validation process. The performance of the models was then evaluated in terms of accuracy in prediction, computational efficiency, utilization of resources, and statistical measures of significance. Findings showed that Random Forest achieved the highest cross-validation score of 1.0000, and Extra Trees managed to obtain the same score of 1.0000. Random Forest had nearly zero discrepancy in prediction, while XGBoost had a score of 0.9999 while RMSE approximating 0.7843. In addition, the Entropy Weight Method (EWM) was applied in the framework to automatically measure the level of importance of cost, latency, and availability rather than using fixed weights. The proposed approach enables data-driven cloud selection while considering both workload requirements and cost-performance factors.
Download PDF