Task Failure Prediction Using Machine Learning Techniques in the Google Cluster Trace Cloud Computing Environment

Abstract

Cloud computing has grown into a critical technology by enabling ground-breaking capabilities for Internet-dependent computer platforms and software applications. As cloud computing systems continue to expand and develop, the need for a more guaranteed, reliant service, and an early task execution status from Cloud Service Providers (CSP) is vital. Additionally, efficient prediction of task failure significantly improves the running time as well as resource utilization in cloud computing. Task failure forecasting in the cloud is regarded as a challenging task based on the literature review conducted in this study. To address these issues, the goal of this study aimed to create fast machine learning approaches for reliably predicting task failure in cloud computing and analyzing their performance using multiple assessment criteria. The Google cluster dataset was used in this study, coupled with Artificial Neural Network (ANN), Support Vector Machine (SVM), and a stacking ensemble method, to forecast job failure in a cloud computing context. The results show that the proposed models can predict the failed tasks both effectively and efficiently. The stacking ensemble outperformed the experimented models, reaching a 99.8%. The suggested paradigm could greatly benefit cloud service providers by decreasing wasted resources and costs associated with task failures.