报告题目：海量非结构化日志的深度分析

报告时间：11月16日 下午15:30

报告地点：信息新楼D507

报告摘要：

System event logs have been frequently used as a valuable resource in data-driven approaches to enhance system health and stability. A typical procedure in system log analytics is to first parse unstructured logs, and then apply data analysis on the resulting structured data. Previous work on parsing system event logs focused on offline, batch processing of raw log files. But increasingly, applications demand online monitoring and processing. We propose an online streaming system Spell, which utilizes a longest common subsequence based approach, to parse system event logs. We show how to dynamically extract log patterns from incoming logs and how to maintain a set of discovered message types in streaming fashion. We also utilize deep-learning based methods to automatically learn useful patterns and models from the underlying log messages. We then use these models to perform online monitoring and anomaly detection. Evaluation results on large real system logs demonstrate that even compared with the offline alternatives, our system shows its superiority in terms of both efficiency and effectiveness.

个人简历：李飞飞博士现任阿里巴巴副总裁，达摩院数据库首席科学家，负责达摩院数据库实验室，以及平台技术群下的数据库事业部和存储技术事业部。加入阿里巴巴之前是美国犹他大学计算机系的终身正教授。他的研究方向是数据库系统，大数据管理理论及系统设计开发，以及云数据管理的安全性。李飞飞博士获得了美国自然科学基金的Caeer Award，美国惠普公司的Innovation Research Program Award，美国谷歌公司的Faculty Award，美国Visa公司的Faculty Research Award。他的研究成果获得了IEEE ICDE 2004 最佳论文奖，IEEE ICDE 2014 10年最有影响力奖，ACM SIGMOD 2015最佳系统演示奖，ACM SIGMOD 2016最佳论文奖， ACM SIGMOD 2017研究亮点奖。他的研究获得了美国自然科学基金以及其他机构和公司的广泛资助，他是VLDB 2014和SIGMOD 2018的演示程序主席，SIGMOD 2014的大会主席，ICDE 2014，SIGMOD 2015，SIGMOD 2019的技术领域程序主席，VLDB 2019和ICDE 2019的博士论坛主席，IEEE TKDE，ACM TODS，Springer DAPD编委会成员。他也是年度SIGMOD Jim Gray最佳博士论文奖评选委员会委员。