Intelligent Log Analytics for Predicting Future Run-Time Issues

用于预测未来运行时问题的智能日志分析

• 批准号: 543528-2019
• 负责人: Zou, Ying
• 金额: $ 4.95万
• 依托单位: Queen's University
• 依托单位国家: 加拿大
• 项目类别: Collaborative Research and Development Grants
• 财政年份: 2021
• 资助国家: 加拿大
• 起止时间: 2021-01-01 至 2022-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=720731
• 关键词: Intelligent; Log; Analytics; Predicting; Future

Mainframe platforms play a central role in supporting mission-critical services that are used by millions of users, such as payroll services, payment services and customer ordering processing services. During the software execution, applications on mainframe platforms are continuously monitored to check the performance and reliability of the applications by tracing a set of key performance metrics and recording the results of the metrics into execution logs. In the current state of practices, the analysis of the execution logs is limited to understanding the already happened events (e.g., crashes and performance bottlenecks). In this proposed project, we plan to apply machine learning techniques to predict future crashes and performance issues by analyzing the historical execution logs. Moreover, we will link the run-time problems with the development to predict the possible run-time failures (e.g., performance bottlenecks) during the development phrase. We will design and develop techniques to improve software performance (e.g., CPU usage, memory usage) in the development stage. Our techniques will give developers early warnings for performance degradation and fix the possible defects before product delivery. The results of the project aim to provide actionable information for the developers to make informed decisions, improve their productivity and enhance the quality and performance of mainframe software systems. The project will train 8 HQPs (i.e., 2PhD, 3MSc, 2BSc and 1RA) in an important and critical topic for applying machine learning and artificial intelligence to the software development operations.

大型机平台在支持数百万用户使用的关键任务服务（例如薪资服务、支付服务和客户订单处理服务）方面发挥着核心作用。在软件执行过程中，对大型机平台上的应用程序进行持续监控，通过跟踪一组关键性能指标并将指标结果记录到执行日志中来检查应用程序的性能和可靠性。在当前的实践中，执行日志的分析仅限于了解已经发生的事件（例如崩溃和性能瓶颈）。在这个拟议的项目中，我们计划应用机器学习技术通过分析历史执行日志来预测未来的崩溃和性能问题。此外，我们会将运行时问题与开发联系起来，以预测开发阶段可能出现的运行时故障（例如性能瓶颈）。我们将在开发阶段设计和开发提高软件性能（例如CPU使用率、内存使用率）的技术。我们的技术将为开发人员提供性能下降的早期警告，并在产品交付之前修复可能的缺陷。该项目的结果旨在为开发人员提供可操作的信息，以做出明智的决策，提高他们的生产力并提高大型机软件系统的质量和性能。 该项目将培训 8 名 HQP（即 2PhD、3MSc、2BSc 和 1RA），培训内容涉及将机器学习和人工智能应用于软件开发操作的重要且关键的主题。