Automatic Detection of Performance Bottlenecks Using a Case-Based Reasoning Approach
Khalil Shihab, Haider Ali Ramadhan · Journal of Intelligent Systems · 2001
Although the process of detecting and resolving performance bottlenecks of computer systems is time-consuming and complex, it has recently become more important and popular issue because of the increase in the complexity and the diversity of these systems.Traditionally, this process has been implemented using rule-based systems.We, however, noted that case-based reasoning is more efficient and of great benefit in this area.This is mainly because non-expert users of performance-modeling or monitoring software, both in preparing inputs and interpreting outputs, tend to use old cases by associating computer systems that reveal similar performance characteristics.It is also because rule-based systems are less modular than case-based systems in that adding (or removing) a new case in a case-based system does not affect any existing case.In this work, we present an application of our integrated approach to the diagnosis of bottlenecks in computer systems in order to provide the necessary remedies to achieve acceptable performance.The underlying technique uses the performance monitor values to find the most possible assignment to the workload and system parameters.It is based on case-based reasoning utilizing fuzzy set concepts to convert the quantitative attributes into qualitative terms for indexing and retrieval.The approach also shows that fuzzy indexing and retrieval are useful in the computer performance domain.