Matching Computing Requirements of Stochastic Optimization Models and Cloud Computing Resources

Cloud computing offers scalable solutions for scientific computing, but efficiently allocating resources for stochastic optimization models remains challenging. This work uses real execution data from an energy sector company to develop machine learning models that predict execution time based on algorithm parameters and cloud infrastructure configurations. To optimize resource usage, we propose a utility-based framework that balances execution time and cloud costs. Our results highlight key factors affecting computational efficiency and provide insights for cost-effective resource provisioning, improving cloud utilization for stochastic optimization applications.