Intelligent Web Services Architecture Evolution Via An Automated Learning-Based Refactoring Framework

Technical debt;Quality of services;Design defects;Architecture evolution;Web service refactoring;Search-based software engineering;Computer and information science;College of Engineering & Computer Science

Wang, HanzhangZhu, Qiang ;

University of Michigan

关键词: Technical debt;    Quality of services;    Design defects;    Architecture evolution;    Web service refactoring;    Search-based software engineering;    Computer and information science;    College of Engineering & Computer Science;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/142810/Wang%20Final%20Dissertation.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Architecture degradation can have fundamental impact on software quality and productivity, resulting in inability to support new features, increasing technical debt and leading to significant losses. While code-level refactoring is widely-studied and well supported by tools, architecture-level refactorings, such as repackaging to group related features into one component, or retrofitting files into patterns, remain to be expensive and risky. Serval domains, such as Web services, heavily depend on complex architectures todesign andimplement interface-level operations, provided by several companies such as FedEx, eBay, Google, Yahoo and PayPal, to the end-users. The objectives of this work are: (1) to advance our ability to support complex architecture refactoring by explicitly defining Web service anti-patterns at various levels of abstraction, (2) to enablecomplexrefactorings by learning from user feedback and creating reusable/personalized refactoring strategies to augment intelligent designers’ interaction that will guide low-level refactoring automation with high-level abstractions, and (3) to enable intelligent architecture evolution by detecting, quantifying, prioritizing, fixing and predicting design technical debts. We proposed various approaches and tools based on intelligent computational search techniques for (a) predicting and detecting multi-level Web services antipatterns, (b) creating an interactive refactoring framework that integrates refactoring path recommendation, design-level human abstraction, and code-level refactoring automation with user feedback using interactive mutli-objective search, and (c) automatically learning reusable and personalized refactoring strategies for Web services by abstracting recurring refactoring patterns from Web service releases. Based on empirical validations performed on both large open source and industrial services from multiple providers (eBay, Amazon, FedEx and Yahoo), we found that the proposed approaches advance our understanding of the correlation and mutual impact between service antipatterns at different levels, revealing when, where and how architecture-level anti-patterns the quality of services. The interactive refactoring framework enables, based on several controlled experiments, human-based, domain-specific abstraction and high-level design to guide automated code-level atomic refactoring steps for services decompositions. The reusable refactoring strategy packages recurring refactoring activities into automatable units, improving refactoring path recommendation and further reducing time-consuming and error-prone human intervention.