Machine Learning Based Techniques for Workload Prediction in Serverless Environments

This paper proposes a novel approach for predicting serverless workloads using machine learning techniques. The research highlights the importance of accurate workload prediction in serverless computing and provides valuable insights into the potential benefits of machine learning in this context. The proposed approach utilises machine learning models to analyse historical data on serverless workloads and various features related to system behaviour to identify patterns and trends and predict the expected resource requirements for future serverless workloads. The study includes multiple machine-learning techniques such as LSTM, ARIMA and VAR and time-series analysis to accurately predict workload patterns and resource demands to determine which is best suited for predicting serverless workloads. The performance evaluation uses real-world serverless workload data, demonstrating its effectiveness in accurately predicting resource needs and optimising resource allocation. This research's findings could improve the efficiency and cost-effectiveness of serverless computing by enabling proactive resource allocation and optimisation.