ОПТИМІЗАЦІЯ ПРОДУКТИВНОСТІ ХМАРНИХ СЕРВІСІВ: МЕТОДИ ТА ЇХ ЕФЕКТИВНІСТЬ

Relevance. Optimizing the performance of cloud services becomes critically important in 2025–2026 due to the exponential growth of data volumes processed within digital ecosystems and the need for rapid adaptation to dynamic workloads driven by the adoption of artificial intelligence (AI) and IoT devices. According to forecasts, the global cloud computing market is growing by 20– 25% annually, and in Ukraine this trend is reinforced by government initiatives such as the National AI Development Strategy through 2025, which identifies cloud services as key infrastructure. However, frequent incidents related to overloads, costs, and security— especially under conditions of energy shortages and geopolitical risks—indicate the fragmented nature of current optimization practices. In Ukrainian business and public administration, where cloud technologies are deployed for digitalization (e.g., e-government and fintech systems), inefficient resource management leads to losses of up to 30–40% of IT budgets. This makes the topic strategically important for ensuring competitiveness, sustainable development, and resilience to cyber threats in multi-cloud environments. The object of the study is modern cloud services, including IaaS, PaaS, and SaaS models provided by platforms such as AWS, Google Cloud, Azure, and local Ukrainian providers, with a focus on performance under variable workloads. Particular attention is paid to the Ukrainian market, where hybrid cloud solutions are combined with local infrastructure to overcome limitations in internet bandwidth and regulatory barriers. The research covers key aspects such as dynamic resource allocation, load balancing, automatic scaling, and AI integration for demand forecasting. The aim is to systematize optimization methods, identify challenges related to scalability, security, and energy efficiency, and develop recommendations for implementation in business (fintech, e-commerce) and the public sector (egovernance, critical infrastructure), taking into account Ukraine’s specific context. Methodology. The article applies a comprehensive approach that includes a comparative analysis of modern optimization methods based on reanalysis data from cloud providers (AWS Auto Scaling, Google Cloud Autoscaler, Azure Autoscale) and implementation case studies from 2025. An expert risk assessment using a factor model is employed to classify causes of inefficiency (human factors, overloads, weak encryption), along with mathematical optimization modeling using SLO/SLA metrics and linear programming to balance costs and performance. Additionally, the study analyzes international standards (ISO 27001 for security, Green Cloud for energy efficiency) and regional Ukrainian data, including numerical experiments with AI-based forecasting algorithms (machine learning for autoscaling). Quantitative assessments are based on statistics such as a 30% cost reduction through AI automation, monitoring via Google Cloud Monitoring, and data center energy consumption modeling. Results. The study confirms that key optimization methods include dynamic resource allocation with AI-based forecasting, which reduces over-provisioning by 25–40%, and automatic scaling that responds to peak loads within seconds, as demonstrated by the Pinterest case on AWS (30% cost reduction). Load balancing and serverless architectures increase performance by 38–60% by eliminating downtime. Challenges include security issues (misconfigured IAM, blind spots in ephemeral resources), energy efficiency (a 40% increase in consumption due to AI), and scaling constraints in Ukraine (limited infrastructure, shadow IT). Provider comparisons show the advantages of hybrid models: Kubernetes orchestration in multi-cloud environments reduces latency by 40%, while neural network quantization cuts computational costs by 40%. In the Ukrainian context, uneven adoption is observed: businesses save on administrative staff, while the public sector suffers from energy shortages. Conclusions. A comprehensive optimization approach is proposed, including AI integration for predictive scaling, standardization of BRM-like procedures for cloud environments (adaptive risk management protocols), and harmonization with local data centers and renewable energy sources to reduce CO₂ emissions by 30%. Recommendations for Ukraine include implementing autoscaling down during off-hours, conducting security testing (CSPM, SIEM), monitoring SLO/SLA metrics, and providing training for AI-driven resource management. The feasibility of hybrid solutions for the sustainable development of the digital ecosystem is demonstrated, ensuring reliability, cost efficiency, and competitiveness in business and public administration, with the potential to reduce costs by 30–50% and increase productivity. This approach lays the foundation for workload forecasting and risk minimization in 2026 and beyond.