Performance problem determination using combined dependency analysis for reliable system

Performance problems such as Service Level Objective violations are determined by specific approach are particularly important since it can significantly improve system reliability. In this paper, we propose an improved performance problem determination mechanism by using combined dependency analysis. Using the proposed preprocessing in prior research, a compact model is created hierarchically. We use temporal inference which associates time with the created models and occurring symptoms, then post symptom into created model to provide probabilistic reasoning. The results of dependency reasoning are combined with the closeness evaluation between symptom and created model. Combined inference is applied according to the characteristic of environment knowledge. Using the improved combined approach enables us to extend problem localization in various situations and it provides much accurate and efficient problem determination technique for reliable system. Performance problem determination using the combined dependency analysis is illustrated to prove the availability and accuracy for improving system reliability.