A fault detection and diagnosis approach for multi-tier application in cloud computing

Ensuring the availability of cloud computing services always concerns both service providers and end users. Therefore, the system always needs precautions for unexpected cases. Accordingly, cloud computing services must be capable of identifying faults and behaving appropriately when it is abnormal to ensure the smoothness as well as the service quality. In this study, we propose a fault detection method for multi-tier web application in cloud computing deployment environment based on the Fuzzy One-class support vector machine and Exponentially Weighted Moving Average method. And then, the suspicious metrics are located by using feature selection method which based on Random Forest algorithm. To evaluate our approach, a multi-tier application is deployed by a transnational web e-Commerce benchmark by using TPC-W (TPC Benchmark™ W, simulates the activities of a business oriented transaction web server in a controlled internet commerce environment) in private cloud and then it is injected typical faults. The effectiveness of the fault detection and diagnosis are demonstrated in experiment results.