Abstract
Context: Software Performance Antipatterns (SPAs) research has focused on algorithms for their characterization, detection, and solution. Existing algorithms are based on the analysis of runtime behavior to detect trends on several monitored variables, such as system response time and CPU utilization. However, the lack of computationally efficient methods currently limits their integration into modern agile practices to detect SPAs in large scale systems.
Objective: In this paper, we extended our previously proposed approach for the automated SPA characterization and detection designed to support continuous integration/delivery/deployment (CI/CDD) pipelines, with the goal of addressing the lack of computationally efficient algorithms.
Method: We introduce a machine learning-based approach to improve the detection of SPA and interpretation of approach's results. The approach is complemented with a simulation-based methodology to analyze different architectural alternatives and measure the precision and recall of our approach. Our approach includes SPA statistical characterization using a multivariate analysis of load testing experimental results to identify the services that have the largest impact on system scalability.
Results: To show the effectiveness of our approach, we have applied it to a large complex telecom system at Ericsson. We have built a simulation model of the Ericsson system and we have evaluated the introduced methodology by using simulation-based SPA injection. For this system, we are able to automatically identify the top five services that represent scalability choke points. We applied two machine learning algorithms for the automated detection of SPA.
Conclusion: We contributed to the state-of-the-art by introducing a novel approach to support computationally efficient SPA characterization and detection that has been applied to a large complex system using performance testing data. We have compared the computational efficiency of the proposed approach with state-of-the-art heuristics. We have found that the approach introduced in this paper grows linearly, which is a significant improvement over existing techniques.