SHF: Medium: Collaborative Research: Finding and Fixing Architectural Hotspots: An Economics-Based Decision Support Approach

最近的研究表明，在源代码文件和软件架构决策中，错误倾向和变更倾向之间存在很强的相关性。也就是说，即使一个软件系统可能有数百个有bug的文件，这些文件总是形成几个架构上相连的组：架构热点。热点显示了在源文件之间传播错误的体系结构缺陷。这种现象已经在许多项目中被观察到，包括开源项目和工业项目，而不考虑它们的领域、年龄或编程语言。这意味着，如果不解决导致这些错误传播的体系结构问题，就不可能减少复杂软件系统中的错误或变更率。这里提出的研究目标是指导高维护体系结构问题的识别，定量地描述它们在软件质量和生产力方面的后果，并创建业务案例来证明它们的重构。最终目标是通过战略性架构改进来减少长期的软件维护成本。这项研究的关键是自动提取架构热点，并根据增加的bug修复工作或降低的交付功能的能力来量化它们的经济后果。这种量化包括构建模型，利用软件项目中广泛可用的信息——关于bug、变更和提交——以便架构师可以计划对热点的重构，并自信地估计此类重构的成本和收益。这项研究将通过pi广泛的国内和国际学术和工业合作产生直接影响。它将从根本上改变软件缺陷被发现、检查和处理的方式：而不是检查数百个有缺陷的文件，每个文件都是孤立的，分析人员只需要检查被建议的方法检测到的几个体系结构热点，通过移除它们的体系结构根源同时修复大量缺陷，从而在维护成本方面提供了大量的节省。建议的决策支持方法通过为软件架构决策的定价和风险分析提供经验基础，具有改变软件行业管理的潜力。本文提出的体系结构热点检测方法将影响众多软件工程研究领域，并为软件体系结构和设计分析的教学提供工具支持，将对软件设计教育产生重大影响。

Recent research has revealed strong correlations between error-proneness and change-proneness in source code files and software architecture decisions. That is, even though a software system may have hundreds of buggy files, these files always form just a few architecturally connected groups: architecture hotspots. Hotspots exhibit architectural flaws that propagate errors among source files. This phenomenon has been observed over numerous projects, both open source and industrial, regardless of their domain, age, or programming language. The implication is that it is impossible to reduce error or change rates in complex software systems without fixing the architecture problems that cause these errors to propagate. The objective of the research proposed here is to guide the identification of high-maintenance architecture problems, quantitatively characterize their consequences in terms of software quality and productivity, and create business cases to justify their refactoring. The end goal is to reduce long-term software maintenance costs though strategic architecture improvement.The key to this research is to automatically extract architecture hotspots, and to quantify their economic consequences in terms of increased bug-fixing effort or reduced ability to deliver features. This quantification involves building models that leverage information broadly available in software projects--on bugs, changes, and commits--so that an architect can plan refactorings to the hotspots and confidently estimate the costs and benefits of such refactorings. This research will produce direct impact through the PIs' extensive national and international academic and industrial collaborations. It will fundamentally change how software defects are discovered, examined, and handled: instead of examining hundreds of defective files, each one in isolation, the analyst only needs to examine a few architecture hotspots detected by the proposed approach, fixing numerous defects simultaneously by removing their architecture roots, thus providing substantial savings in maintenance costs. The proposed decision-support approach has the potential to change the management of the software industry by providing an empirical basis for the pricing and risk analysis of software architecture decisions. The proposed architecture hotspot detection approach will influence numerous software engineering research areas, and will have significant impact on software design education by providing tool-support for the teaching of software architecture and design analysis.