Lessons Learned: Automated Event Recognition in Distributed Data Environments

摘要

We investigated issues with recognizing events when monitoring large amounts of live data coming from distributed systems. We did this by repotting on a system that was deployed in NASA/JSC's Water Research Facility. The system included complex event recognition software that recognized significant events and anomalies when monitoring the Water Recovery System. We share our experiences and lessons learned after running the system for a year and a half. We discuss the issues that were brought about by operating in a distributed data environment. We believe that these issues will need to be addressed by any system that performs complex event recognition in such an environment. This is partly due to the fact that recognizing events is sequential by nature, and operating in a distributed data environment is parallel by nature. In this paper we discuss our solutions to these issues and point out areas that require further research for future growth of this technology.