A prediction model for improving virtual machine live migration performance in cloud computing using artificial intelligence techniques

• Published in: International journal of computers & applications 2024-12, Vol.46 (12), p.1069-1087
• Main Authors: Alubaidan, Haya A., Aljameel, Sumayh S.
• Format: Article
• Language: English
• Subjects: Artificial intelligence; Cloud computing; Data centers; Deep learning; Downtime; live migration; Machine learning; Migration; Prediction models; Predictions; Regression models; Virtual environments; Virtualization
• ISSN: 1206-212X; 1925-7074
• DOI: 10.1080/1206212X.2024.2409988

Over recent years, virtualization has worked as the powerhouse of the data centers. To positively influence datacenter utilization, power consumption, and management, live migration presents a technique which must be employed. Live migration refers to the method whereby an online virtual machine or an application is moved from one physical server to another without shutting down the client or the program. However, a migration's qualities, which define service quality, have to be evaluated prior to a migration. In this paper, we propose a novel approach using predictive machine learning and deep learning models to forecast key metrics such as total migration time, downtime, and total data transferred for two migration types: pre-copy migration and post-copy migration. It builds on methods from regression machine learning and deep learning, and hyperparameter optimization, and feature engineering to push past reported solutions. The novelty of this work is the development of the hybrid model that includes the positive aspects from both of the machine learning and the deep learning with an emphasis on the increased accuracy of the model's predictions. Concretely, the proposed deep learning framework based on GRU provided outstanding performance with 99.6% in total migration time where the transferred data GRU achieved 99.9% Moreover, in Down time target GRU reached 98.3%. This outperforms the other machine learning model and prior studies to become the best model for live migration prediction.