Auto-Scaling Cloud Resources Forecasting Approach Based On Time Series

Auto-scaling is the mechanism that is used to monitor and permanently control resources of cloud Computing by increasing or decreasing allocated resources in accordance with the expected time of the machine workload, the future workload can be predicated by machine learning (ML) algorithms. The main purpose of using Autoscaling Machine resources in cloud services solutions is to keep performance and remain operational cost low by reducing resources utilization. In this paper, we propose a prediction model that is based on three approaches: Autoregressive Integrated Moving Average (ARIMA), Vector Autoregression (VAR), Exponential Smoothing (ES), and Long Short-Term Memory Recurrent Neural Network (LSTM). Series of experiments and Implementation results show that this model can forecast machine workload with accuracy up to 97.0 %. The key contribution of this paper is to select the best prediction mechanism for machine workload.