As businesses attempt to boost flexibility and cut costs, cloud computing is becoming more popular. Despite the fact that the big cloud service providers use a pay-as-you-go pricing model and allow clients to scale up and down easily, there is still space for improvement. Workload, measured in terms of CPU utilization, fluctuates frequently, resulting in excessive costs and environmental damage for businesses. The goal of this paper is to use a long short-term memory machine learning model to forecast future CPU consumption. Companies can scale their capacity just in time and minimize excessive costs and environmental damage by estimating utilization up to 5 minutes in advance over a 30-day period. The analysis is split into two sections. The first section compares the performance of the LSTM model to a state-of-the- art model when predicting one step at a time. The second section examines the LSTM’s accuracy when making predictions up to 5 minutes in advance over a 30-day period. To determine the optimal LSTM for the prediction, we compared three distinct LSTMs. To sum up, the study found that LSTM could be a beneficial model for lowering costs and eliminating unnecessary environmental effects for commercial applications hosted on the cloud.

Cloud Computing, Machine Learning, LSTM, Prediction.