SLA Violation Prediction In Cloud Computing: A Machine Learning Perspective

Service level agreement (SLA) is an essential part of cloud systems to ensure maximum availability of services for customers. With a violation of SLA, the provider has to pay penalties. Thus, being able to predict SLA violations favors both the customers and the providers. In this paper, we explore two machine learning models: Naive Bayes and Random Forest Classifiers to predict SLA violations. Since SLA violations are a rare event in the real world (∼ 0.2%), the classification task becomes more challenging. In order to overcome these challenges, we use several re-sampling methods such as Random Over and Under Sampling, SMOTH, NearMiss (1,2,3), One-sided Selection, Neighborhood Cleaning Rule, etc. to re-balance the dataset. We use the Google Cloud Cluster trace as the dataset to examine these different methods. We find that random forests with SMOTE-ENN re-sampling have the best performance among other methods with the accuracy of 0.9988% and F1 score of 0.9980.