AI Based Cloud Failure Detection and Prevention Algorithm

It is crucial to anticipate and avoid cloud breakdowns to guarantee the availability and dependability of cloud services. Machine learning algorithms have shown promise in forecasting and preventing cloud failures by analyzing large datasets of cloud usage traces to identify potential failure points. This paper presents an algorithm incorporating different Machine Learning and Deep Learning models for cloud failure prediction and prevention based on analyzing the Google cluster usage traces dataset. The proposed algorithm, the AI-based Cloud Failure Prevention Algorithm (ACFP), utilizes various features to predict potential failure points. The proposed algorithm used various AI models, the prominent ones are Cat Boost (accuracy, F1 score, Precision, Specificity: 100%), LGB (accuracy, F1 score, Precision, Specificity: 100%), Gradient Boost (accuracy, F1 score, Precision, Specificity: 100%), Random Forest (accuracy, F1 score, Precision, Specificity: 100%), etc. for the precise detection of cloud task failure.After that the ACFP selects one ML/DL model with best failure prediction efficiency and use some parameters like: “resource_utilization”, “event_type”, and “scheduler” to generate some suggestion based values for the cloud so that the cloud implements that suggestion to prevent cloud failure. The experimental results show how well the algorithm predicts and averts cloud breakdowns, highlighting its potential to raise the reliability and availability of cloud services. The comparison of the failure rate of the cloud by predicting job failures with and without the ACFP is 2.67% and 22.76%, respectively. This algorithm can raise the quality of cloud services and reduce the risks of cloud failures.