Systematic and Automated Software Migration via Model-Driven Engineering

点击翻译按钮获取中文摘要

Background Software migration is the process of transforming software applications from one operating environment to another. It impacts on all software intensive industries, including enterprise systems, avionics, automotive - where operating systems, hardware, languages or frameworks may change. Migration is expensive. The Cooperative Bank UK lost GBP 1.5B on a failed migration project. Netscape spent 2 years on a migration which led to loss of market share and ultimately the collapse of the company. Many real-world migration problems are handled manually, with intensive developer effort and little automated support. Software migration thus involves time-consuming, repetitive and error-prone tasks. The work this proposal describes will produce novel tools and theories, based on Model-Driven Engineering (MDE), to support engineers engaged in software migration. It provides tools to support engineers in automating migration tasks and supporting engineers in predicting what problems may be encountered. Goal The goal of the research is to develop and evaluate techniques and tools for software migration across changes to application operating environments. This fits with my long-term research focusing on automating repetitive & error-prone tasks using MDE. My research will consider a broad set of migration scenarios triggered by hardware or software changes, as well as a new method for software migration that allows early analysis of potential violations of timing properties. Challenges (i) Migration scenarios are diverse and are generally tackled individually. (ii) Automated support exists for specific migration problems but not for an end-to-end process. (iii) There is little support for prediction of problems prior to migration. (iv) Guidance on when migration can be partially supported is largely non-existent. (v) The influence of programming language on migration problems related to timing is poorly understood. Research Outcomes and Directions I will build a novel catalogue of recurring problems in migration across a range of different operating environments, and develop innovative executable model transformations to address them. I will develop a new methodology identifying migration 'hotspots', as well as novel mechanisms for prediction of timing problems. I will develop new guidance on when partial migration may be possible/beneficial. Impact Maintenance is the most expensive part of software engineering; the proposed work will yield productivity improvements that will reduce overall costs when companies must undertake migration. The research will lead to an innovative catalogue of software migration problems and patterns, as well as new mechanisms for predicting the impact of migration on timing characteristics of software. It will strengthen Canada's position as a leader in software migration, both in academia and industry, and will build on strong synergies between Canadian universities and companies working in software maintenance.