FORMALISATION OF REQUIREMENTS FOR INTELLIGENT DECISION SUPPORT SYSTEMS IN CLOUD RESOURCE OPTIMISATION TASKS

Abstract

In the context of modern cloud computing environments, there is an increasing demand for automated resource management systems to ensure optimal performance and cost efficiency. Intelligent decision support systems are a vital part of these mechanisms, providing analysis of dynamic metrics and forecasting of load changes for adaptive scaling.Objective. The objective is to develop a formal model of the requirements for intelligent decision support systems for optimising cloud infrastructure resources.Methodology. This study analyses existing approaches to autoscaling and intelligent solutions for managing cloud environments. System analysis methods were employed to identify categories of requirements and construct an ontological model comprising the following entities: CloudResource, MonitoringAgent, MLModel, ScalingPolicy and DecisionRecord. Experimental validation was performed by implementing a prototype in a Kubernetes environment and using Prometheus to collect metrics.Scientific novelty. This work is scientifically novel due to its comprehensive formalisation of requirements for intelligent decision support systems, considering both functional and non-functional aspects. The functional requirements include monitoring heterogeneous metrics, forecasting mechanisms, generating scaling recommendations and auditing decisions.Conclusion. The results obtained show that the formalisation of requirements is a prerequisite for the creation of reliable and transparent intelligent autoscaling systems in cloud environments. The proposed model strikes an effective balance between performance and infrastructure costs, ensuring the optimal use of resources without compromising service quality.Future work should focus on extending the ontology to cover multi-cloud and serverless scenarios, developing methods for rapidly updating models (concept drift) and improving Explainable AI mechanisms to provide more transparent justification for recommendations.