Analytics and Data Summit 2019

While most data analysis is textual, this presentation discusses the use of Java Map-Reduce to extract the metadata from sonar files collected by UUV (Unmanned Underwater Vehicle) missions. The sheer volume of files generated makes this a good candidate for Hadoop divide and conquer. The geolocation and timestamp information is extracted and loaded into a Hive Database, followed by a Sqoop job to transfer the data to an Oracle Database. Oracle APEX (Application Express) is then used to visualize the information on a Google map. The motivation for this effort comes from a prior project that extracted event information from Windows Event logs (using Java Map-Reduce). In addition the depth information is also used with basic AI classification algorithms in Hadoop/Spark for prediction.

This session focuses on how Oracle is applying various Machine Learning technologies to the area of AI Opts to prevent performance issues and maintain availability. Details about the architectures and algorithms used for this Autonomous Health initiative will be covered for both cloud and on-premise database deployments. Specific operational use cases will be covered detailing the strategies and algorithms used to proactively detect performance degradation through the phases of detection, root-cause analysis and targeted corrective actions. The first section focuses on detection based upon data from the operating system and software layers and how normality can be determined and deviations detected in realtime with the goal of preserving session, database, and infrastructure availability. The second section focuses on detection using log and trace file data were the analysis is based on anomalous events or log record signatures. The final section will cover integration into DevOps processes.