A carregar...
Publicação
Energy Consumption and Resources Prediction for Container-Based Solutions
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 3.17 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
As cloud and fog infrastructures continue to grow, the increasing volume of data shared across these systems presents significant challenges in managing energy consumption efficiently. Addressing these challenges is essential to meet Green IT standards and ensure sustainable computing practices. This thesis introduces a novel approach to energy consumption prediction, providing a non-intrusive method for assessing the energy usage of containerized environments. The primary focus of this work is on developing a model capable of accurately predicting the energy consumption of container-based solutions using their expected command as the main feature. Using an extensive dataset encompassing metrics such as voltage, current, power, frequency, energy, and power factor, various regression models were evaluated to determine the most effective approach. The selected model leverages these measurements to forecast future energy consumption over extended periods, offering actionable insights into energy usage patterns. This predictive framework empowers users to optimize resource allocation, reduce energy costs, and enhance IT equipment efficiency. By applying advanced feature importance analysis and hyperparameter optimization, the study highlights critical features influencing energy consumption, such as CPU and memory usage rates. The results demonstrate the potential for organizations to proactively manage energy consumption strategies in dynamic cloud and fog environments, supporting both operational efficiency and environmental sustainability.
Descrição
Tese de Mestrado, Informática, 2025, Universidade de Lisboa, Faculdade de Ciências
Palavras-chave
Prediction Model Regression Energy Consumption Resource Optimization Containers