INFORMATION TECHNOLOGY FOR OPTIMAL SERVICE PLACEMENT PREDICTION IN A MULTICLOUD ENVIRONMENT USING MACHINE LEARNING

Abstract

The relevance of the work is due to the need to improve the efficiency of service distribution management in multi-cloud infrastructures, where optimal service placement directly affects latency, performance, reliability, and rational use of resources. The object of the study is the process of placing cloud services in a multi-provider environment. The subject of the study includes machine learning methods and algorithms that are used to predict optimal decisions for placing cloud services in a multi-provider environment based on measured performance indicators. The purpose of the study is to develop and evaluate models for predicting optimal placement of cloud services in a multi-provider environment using historical data on latency, response time, and load balancing efficiency. The work uses an open dataset, the Multi-Cloud Service Composition Dataset, which contains characteristics of services from AWS, Azure, Google Cloud, and IBM providers. Six machine learning algorithms implemented using the Python programming language and the Scikit-learn library were used for prediction. The obtained results showed that models based on Gradient Boosting and Naive Bayes provide the highest consistency of the metrics Accuracy, Precision, Recall and F1-score, reaching values of about 0.97–0.98, which confirms their suitability for the tasks of optimizing the placement of cloud services in a multi-cloud environment. Other developed models demonstrated lower stability of results, which limits their application in real conditions. The conclusions substantiate the possibility of using machine learning methods and algorithms to build adaptive load management systems in multi-cloud environments, and also identify prospects for expanding the proposed information technology by including additional parameters, such as energy consumption, computing cost and fault tolerance.