Unifying Service Oriented Technologies for the Specification and Detection of Their Antipatterns

Sammanfattning: Service-based Systems (SBSs) are developed on top of diverse emerging Service-Oriented Architecture (SOA) technologies and architectural choices, including SOAP Web services, SCA (Service Component Architecture), and REST. Yet, like any other complex systems, SBSs are subject to change. The changes can be functional or non-functional and can be at design or implementation-level. Such changes may degrade the quality of design and quality of service (QoS) of the services in SBSs by introducing poor solutions—service antipatterns. The presence of service antipatterns in SBSs may hinder the future maintenance and evolution of SBSs. Assessing the quality of design and QoS of SBSs through the detection of service antipatterns may ease their maintenance and evolution. With a few commonalities among various SBSs implementation technologies, they vary in their (1) building blocks, (2) composition styles, (3) development methodology, and (4) communication or client interaction styles; which pose challenges to analyse them in a unique manner. However, the current literature lacks a unified approach for evaluating the design quality and QoS of SBSs.To address this need, this dissertation presents an abstraction unifying three SBSs technologies: Web services, SCA, and REST. Using this abstraction, it describes a unified approach, SODA (Service Oriented Detection for Antipatterns), supported by a framework, SOFA (Service Oriented Framework for Antipatterns), for assessing the design and QoS of SBSs. Using the SODA approach, we define detection rules for 31 service antipatterns and 10 service patterns independent of their technologies. Based on these rules, we automatically generate (for SCA and Web services) or implement (for REST) their detection algorithms. We apply and validate these algorithms in terms of precision and recall on (1) two SCA systems, (2) more than 120 SOAP Web services, and (3) a set of 15 widely-used RESTful services, including Facebook, Twitter, and YouTube.The detection results provide evidence of the presence of service antipatterns in SBSs. The reported detection accuracy exhibits a high precision and recall and an acceptable detection performance in terms of detection time. Our SODA approach and the underlying tool can help practitioners to evaluate their SBSs, which may result in an SBS (1) with improvedquality of design and easy maintenance and (2) with improved QoS for the end-users than an SBS with antipatterns.