Quality and Developer Productivity Enhancements for Cloud-Native Applications via Fault Analysis & Localization with Machine Learning

该项目研究机器学习和数据挖掘技术，以基于错误形成的代码和系统配置设置导致的运行时行为来识别和定位云本地微服务应用程序的故障。它还调查了推荐系统的开发，以帮助开发人员在识别和定位这些运行时错误后减轻这些错误。这些运行时故障通常将自身暴露为应用程序中的性能降级故障，而此类故障不会作为软件故障出现。由于系统中不同的组件、系统和编译器设置以及系统中组件之间的低效交互，使得故障的识别和定位变得困难。将研究一种机器学习方法来检测、识别和定位故障。它的有效性将基于其将运行时跟踪错误与应用程序代码结构、配置参数和/或组件交互行为相关联的能力进行评估。作为一个推荐系统，我们设想，一旦确定错误，我们将搜索针对错误类型建议的软件开发指南，并将此信息呈现给开发人员。建议可以以两种模式应用：作为开发人员编码工作流程的一部分，在集成开发环境(IDE)中交互应用，或作为生成潜在问题报告和建议修复的独立工具。行业合作伙伴支持基于Java的云微服务的开放开发平台，并且不需要此类工具。目前，还没有这样的工具可以检查运行时故障并向开发人员建议对云本地微服务应用程序进行更改。

This project investigates machine learning and data mining techniques to identify and locate faults of cloud-nativemicro-service applications based on runtime behaviours caused by poorly formed code and system configuration settings. It also investigates the development of a recommendation system to help developers mitigate these run-time faults once they have been identified and located. These run time faults typically expose themselves as performance degradation faults in the application and assuch do not surface as software failures. This makes it difficult to identify and localize the fault as it could becaused from different components in the system, system and compiler settings, as well as inefficientinteractions among the components in the system.A machine learning approach will be investigated for the detection, identification, and localization of thefaults. Its effectiveness will be evaluated based on its ability to correlate runtime trace faults to applicationcode structure, configuration parameters, and/or component interaction behaviours.As a recommendation system, we envision that once the fault is identified we would search for therecommended software development guides for the fault types and present this information to the developer.The recommendation could be applied in two modes: interactively in an Integrated Development Environment(IDE) as part of the developer's coding workflow or as a standalone tool that produces a report of potentialissues and suggested fixes.The industrial partner supports an open development platform for Java-based cloud micro-services and are inneed of such tools. Currently, no such tools are available that can examine runtime faults and recommendchanges to the developer for cloud-native micro-service applications.