Analysis of Job Failure Prediction in a Cloud Environment by Applying Machine Learning Techniques

Abstract

Cloud services are the on-demand availability of resources like storage, data, and computing power. Nowadays, cloud computing and storage systems are continuing to expand; there is an imperative requirement for CSPs (Cloud Service providers) to ensure a reliable and consistent supply of resources to users and businesses in case of any failure. Consequently, large cloud service providers are concentrating on mitigating any losses in a cloud system environment. In this research, we examined the bit brains dataset for job failure prediction, which keeps traces of 3 years of cloud system VMs. The dataset contains data about the resources used in a cloud environment. We proposed the performance of two machine learning algorithms: Logistic-Regression and KNN. The performance of these ML algorithms has been assessed using cross-validation. KNN and Logistic Regression give optimal results with an accuracy of 99% and 95%. Our research shows that using KNN and Logistic Regression increases the detection accuracy of job failures and will relieve cloud-service providers from diminishing future losses in cloud resources. Thus, we believe our approach is feasible and can be transformed to apply in an existing cloud environment.