Automatic Failure Localization and Diagnosis for Cloud Computing Applications

点击翻译按钮获取中文摘要

SAP is a global software company that develops and operates large-scale computing infrastructure and servicesfor enterprise applications. A crucial aspect of SAP's strategy is to offer their clients Software as a Service(SaaS) solutions based on an open-source technology integration layer and a micro-service architecture.Troubleshooting micro-service-based cloud applications is challenging because, while a problem may occur inany of the distributed micro-services required for an application, developers cannot directly inspect theexecution of services in operation. The standard practice for troubleshooting cloud-based applications islimited to analyzing logs written by the various services. As a result, developers must engage in highly manualand ineffective inspections that follow an iterative process of elimination for potential root causes.We propose to investigate how to develop new analyses of distributed system logs to better support failurelocalization and description in micro-service-based cloud applications that use technology relevant to SAP.The project will consist in systematic investigation of the information logged by different types of cloudservices (e.g., interpreter, database management system) and experimentation with new strategies for mappingfailure symptoms detected in logs to concepts relevant to software developers.