In the complex and dynamic environment of cloud computing, the workload prediction of virtual machines has attracted great attention in research areas because the workload of virtual machines plays a significant role in improving the quality of services and reducing energy consumption over the cloud environment. Hence, an effective workload prediction is essential to efficient resource provisioning in cloud computing.
Conventional machine learning-based workload prediction models do not effectively predict the workload, and it was confronted with several challenges due to the high variance and high dimensionality of cloud workloads. To tackle this constraint, an efficient deep learning model is necessary to design effectively and extract and learn the essential representations of workloads from the original high-dimensional workload. Deep learning-based workload prediction models such as Recurrent Neural Network can address the problem of long-term memory dependencies in cloud workload prediction, and it achieves adaptive and accurate prediction for highly-variable workloads.