Workload Forecasting Model for Resource Management in Cloud Data Center

Cloud computing is a framework that can provide services to clients by virtually sharing resources in data centers. Studies suggest that predicting workload traits in advance can help in managing resources efficiently, resulting in saving power, reduced cost, and improved service performance. Accurate workload prediction methods estimate the number of resources to be allocated to each application in cloud data centers. We evaluate multiple machine learning and deep learning (ML and DL) algorithms for prediction of workload patterns. We propose a DL based model to predict workload, which results in improved decision making for better management of resources. Two common evaluation metrics MSE and RMSE, have been employed to analyze to verify suggested workload prediction model. It is found that based on experimental results, the Long Short Term Memory (LSTM) algorithm performed exceptionally well, achieving the lowest RMSE value among all other algorithms tested for workload performance prediction.