Dynamic Resource Optimization in Cloud Computing

Abstract

   Dynamic resource allocation in cloud computing environments is a major challenge that requires accurate prediction and effective optimization to improve resource utilization and reduce costs. Traditional methods often have difficulty in adapting quickly to fluctuating demand, leading to waste or service quality degradation. This study proposed an integrated method for forecasting resource demand (CPU, memory, bandwidth), using two optimization algorithms: Genetic Algorithm and Ant Colony Optimization. A Transformer model helped accurately predict future demand based on historical data and allowed proactive adjustment of resource allocation. Genetic Algorithm and Ant Colony Optimization were then applied to optimize resource allocation across virtual machines. Experiments on real data showed that this method significantly improved prediction accuracy, reduced allocation delay, and optimized resource utilization better than traditional methods such as LSTM or search-based optimization. This finding has offered a promising solution for dynamic resource management in cloud computing.