Web service design defects detection: A bi-level multi-objective approach

Abstract

Abstract Context: Web services frequently evolve to integrate new features, update existing operations and fix errors to meet the new requirements of subscribers. While this evolution is critical, it may have a negative impact on the quality of services (QoS) such as reduced cohesion, increased coupling, poor response time and availability, etc. Thus, the design of services could become hard to maintain and extend in future releases. Recent studies addressed the problem of web service design antipatterns detection, also called design defects, by either manually defining detection rules, as combination of quality metrics, or generating them automatically from a set of defect examples. The manual definition of these rules is time-consuming and difficult due to the subjective nature of design issues, especially to find the right thresholds value. The efficiency of the generated rules, using automated approaches, will depend on the quality of the training set since examples of web services antipatterns are limited. Furthermore, the majority of existing studies for design defects detection for web services are limited to structural information (interface/code static metrics) and they ignore the use of quality of services (QoS) or performance metrics, such as response time and availability, for this detection process or understanding the impact of antipatterns on these QoS attributes. Objective: To address these challenges, we designed a bi-level multi-objective optimization approach to enable the generation of antipattern examples that can improve the efficiency of detection rules. Method: The upper-level generates a set of detection rules as a combination of quality metrics with their threshold values maximizing the coverage of defect examples extracted from several existing web services and artificial ones generated by a lower level. The lower level maximizes the number of generated artificial defects that cannot be detected by the rules of the upper level and minimizes the similarity to well-designed web service examples. The generated detection rules, by our approach, are based on a combination of dynamic QoS attributes and structural information of web service (static interface/code metrics). Results: The statistical analysis of our results, based on a data-set of 662 web services, confirms the efficiency of our approach in detecting web service antipatterns comparing to the current state of the art in terms of precision and recall. Conclusion: The multi-objective search formulation at both levels helped to diversify the generated artificial web service defects which produced better quality of detection rules. Furthermore, the combination of dynamic QoS attributes and structural information of web services improved the efficiency of the generated detection rules.