دانلود On the Value of Quality of Service Attributes for Detecting Bad Design Practices

Abstract

Service-Oriented Architectures (SOAs) success-fully evolve over time to update existing exposed features to the users and fix possible bugs. This evolution process may have a negative impact on the design quality of Web services. Recent studies addressed the problem of Web service antipatterns detection (bad design practices). To the best of our knowledge, these studies focused only on the use of metrics extracted from the implementation details (source code) of the interface and the services. However, the quality of service (QoS) metrics, widely used to evaluate the overall performance, are never used in the context of Web service antipatterns detection. We start, in this work, from the hypothesis that these bad design practices may impact several QoS metrics such as the response time. Furthermore, the source code metrics of services may not be always available. Without the consideration of these QoS metrics, the current detection processes of antipatterns will still lack the integration of symptoms that could be extracted from the usage of services. In this paper, we propose an automated approach to generate Web service defect detection rules that consider not only the code/interface level metrics but also the quality of service attributes. Through multi-objective optimization, the proposed approach generates solutions (detection rules) that maximize the coverage of antipattern examples and minimize the cover-age of well-designed service examples. An empirical validation is performed with eight different common types of Web design defects to evaluate our approach. We compared our results with three other state of the art techniques which are not using QoS metrics. The statistical analysis of the obtained results confirm that our approach outperforms other techniques and generates detection rules that are more meaningful from the services’ user perspective.