INCREASING THE EFFICIENCY OF CLOUD RESOURCE MANAGEMENT WITH THE HELP OF DISTRIBUTED MONITORING

Abstract

The purpose of this work is to develop an architecture for a distributed monitoring system capable of adaptively responding to variable load, partial failures, and network degradation to improve the efficiency of cloud computing resource management.The research methodology is based on the development of a hierarchical architecture that includes monitoring agents with adaptive sampling rates, a fog layer with local ML models for load forecasting, and a central coordinator that performs global analytics and strategic management. During the study, an experimental prototype of the system was implemented based on a containerized environment using Docker, Apache Kafka, Python, Random Forest, and SHAP interpretation. A series of experiments were conducted under five scenarios: from nominal operation to stress conditions with loss of connectivity, overloads, and combined threats.Scientific novelty – for the first time, a combination of agent monitoring with a dynamic sampling rate and ML analytics at the fog level has been implemented to ensure high speed and accuracy of response in decentralized management conditions.The proposed model outperforms traditional centralized approaches in both qualitative and quantitative indicators.Conclusions. According to the results of experiments, it was found that the proposed architecture allows reducing the average system response time to events by 60–70 %, reducing telemetry data losses by 3–4 %, increasing the accuracy of load forecasting by up to 30 %, and reducing the number of erroneous actions in dynamic infrastructure conditions. In addition, the distributed nature of information processing provides a higher level of system stability by reducing fluctuations in resource allocation. The results obtained confirm the feasibility of implementing the proposed architecture in systems that require high reliability, adaptability and scalability – in particular, in edge/fog environments, industrial Internet of Things, smart data centers and critical real-time services.