An LSTM-based Approach for Predicting Resource Utilization in Cloud Computing
Tu Nguyen, Van-Tri Do, Khanh Le, Seungkyu Go, Sunghyun Na, Dukyun Kim, Duc Tran · 2022
Predicting future resource consumption has become a significant issue as large-scale cloud computing centers surpass individual servers in popularity. Public cloud service providers can proactively assign or reallocate resources for cloud services by forecasting resource needs. This research aims to forecast the usage of resources such as the central processing unit, random access memory, and hard disk across both short-term and long-term time scales. In this paper, we propose to use Long Short-Term Memory network (LSTM) with our own approach for resources’ usage prediction in cloud workloads. The proposed approach has been evaluated and compared with other traditional approaches on predicting cloud workloads. The experimental results show that such approach provides more accurate predictions with at least two times lower loss values, measured in terms of median absolute error for both long-term and short-term prediction. This work helps the cloud service provider (CSP) to analyze and predict the workload accordingly to acknowledge over and under provisioning of the cloud resources.