Enhanced virtual machine consolidation : an in-depth look at ANN and BiGRU based virtual machine selection

The proliferation of cloud computing has been transformative across multiple sectors, enabling unprecedented computational capabilities. With the sheer scale of operations in modern data centers, there is an inevitable surge in energy demands. This exponential growth underscores the need for efficient resource allocation strategies that can ensure sustainability without compromising on performance. Recognizing this challenge, numerous strategies have been introduced in recent years for effective virtual machine (VM) distribution and allocation in data centers. Such strategies strive to strike a balance between maximizing resource utilization and mitigating energy consumption. In pursuit of a more accurate and energy-efficient cloud environment, this research delves into the intricacies of dynamic VM consolidation. We introduce a novel model that marries the prowess of deep learning techniques with traditional machine learning. Specifically, our model harnesses the spatial feature extraction capabilities of Artificial Neural Networks (ANN) with Bidirectional Gated Recurrent Units (BiGRU) and enriches it with an attention mechanism, emphasizing the salient features. Utilizing key metrics such as CPU requests, memory requests, average memory consumption, and average CPU utilization, our approach offers a granular insight into server performance, enabling informed VM migration decisions. Empirical validation on a dataset showcases the potential of our model, achieving superior accuracy rates in VM failure prediction. When compared to benchmark strategies, our approach demonstrates improvements in prediction accuracy, underscoring its value in fostering a more resilient and efficient cloud computing ecosystem. The validity of the proposed model was verified using a dataset from Google, and the results demonstrated an improvement in accuracy by approximately 3.5–27% compared to other algorithms proposed in the reference study.