Scholarlink

<p>Building an effective cloud workload prediction system is challenging due to the complex and dynamic nature of cloud computing environments. Accurate prediction of resource utilization, such as CPU and memory loads, is crucial for efficient resource provisioning and management. High-performance machine learning models are needed for cloud workload prediction, as the workloads can exhibit diverse patterns and changes over time. This study proposes a unique ensemble framework that can effectively predict different aspects of cloud workloads. The proposed approach is based on building an ensemble learning-based models, such as XGBoost and LightGBM, to predict CPU and memory utilization in the cloud computing environment. Specifically, a two-stage model was developed, where the first stage predicts the workload status and the second stage predicts the actual CPU and memory usage based on the predicted status. Hyperparameter tuning is employed to optimize the performance of the ensemble models. The results of the proposed approach demonstrate its effectiveness in cloud workload prediction. For CPU utilization, the XGBoost model achieves an R-squared value of 0.97967, which is 25.78% better than the ARIMA baseline. Similarly, for memory utilization, the LightGBM model achieves an R-squared value of 0.949. These results highlight the superiority of the proposed ensemble framework over traditional approaches, enabling more accurate and reliable cloud resource provisioning and management.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Improving cloud resource management: an ensemble learning approach for workload prediction

  • 期刊论文

  • 发表日期：

    2025年7月10日

查看全文
  • Jyoti Bawa,
  • Kuljit Kaur Chahal,
  • Kamaljit Kaur

摘要

Building an effective cloud workload prediction system is challenging due to the complex and dynamic nature of cloud computing environments. Accurate prediction of resource utilization, such as CPU and memory loads, is crucial for efficient resource provisioning and management. High-performance machine learning models are needed for cloud workload prediction, as the workloads can exhibit diverse patterns and changes over time. This study proposes a unique ensemble framework that can effectively predict different aspects of cloud workloads. The proposed approach is based on building an ensemble learning-based models, such as XGBoost and LightGBM, to predict CPU and memory utilization in the cloud computing environment. Specifically, a two-stage model was developed, where the first stage predicts the workload status and the second stage predicts the actual CPU and memory usage based on the predicted status. Hyperparameter tuning is employed to optimize the performance of the ensemble models. The results of the proposed approach demonstrate its effectiveness in cloud workload prediction. For CPU utilization, the XGBoost model achieves an R-squared value of 0.97967, which is 25.78% better than the ARIMA baseline. Similarly, for memory utilization, the LightGBM model achieves an R-squared value of 0.949. These results highlight the superiority of the proposed ensemble framework over traditional approaches, enabling more accurate and reliable cloud resource provisioning and management.