Workload Prediction over Cloud Server using Time Series Data
Mahendra Pratap Yadav, Nisha Pal, Dharmendar Kumar Yadav
Abstract
Analyzing and interpreting the real time data is a challenging task for cloud analysts (e.g. cloud providers) in cur-rent scenario for allocating computing resources in applications. Availability of a massive amount of data for processing over a server, cloud providers use the time series analysis models to analyze it. Based on the analysis, cloud providers allocate cloud resources (e.g., container machines) to manage the workload. Predictive analysis of data is important to identify the future trends and it also enables cloud organizations to act as per the demand of workload. It can be applied in different areas such as stock prices prediction, weather forecasting, and traffic load over the server (cloud computing). Cloud providers use this predictive analysis to avoid different types of losses such as services unavailability, maximum energy consumption and customer's loss. One of the methods to do predictive analysis using time series data is long short-term memory (LSTM). This paper presents a predictive analysis of time series forecasting using deep learning method (LSTM) to predict the future load over servers. The prediction accuracy of LSTM has been measured using three metrics -RMSE, MSE and MAE.