Workload Prediction based on GRU-CNN in Cloud Environment

Workload prediction is crucial to enforce various optimizations of cloud services. Deep learning has recently achieved great success in many fields due to its robust feature representation capabilities. Deep-learning-based workload prediction is no exception. However, the performance is still inadequate for real-world application. This paper proposes an effective combined model of GRU(Gated Recurrent Unit) and CNN(convolutional neural networks). With GRU's strong time-varying dependency capturing and CNN's robust pattern recognition, the proposed model can accurately predict the workload of cloud servers. Experiments show that the proposed GRU-CNN model outperforms other state-of-the-art models regarding prediction accuracy and real-time performance.