Workload and renewable energy prediction in cloud data centers with multi-scale wavelet transformation

In recent years, cloud computing and big data services are widely adopted by large-scale enterprises. The energy consumption of cloud data centers (CDCs) has also increased dramatically. To effectively reduce the harm on the environment, a growing number of CDCs consider renewable energy instead of fossil energy, and concentrate on reducing idle time of servers by forecasting short-term workload demands for proactively provisioning computational resources and balancing server load in advance. However, due to temporal fluctuation in workload demands and renewable energy, it is a huge challenge to precisely predict their short-term trends. This work adopts basic methods in the field of signal processing and proposes a time series prediction method based on multi-scale wavelet transformation. Extensive experiments based on real-life datasets demonstrate that the proposed method achieves higher accuracy than several typical baseline methods.