Leveraging many simple statistical models to adaptively monitor software systems
ISPA'07 Proceedings of the 5th international conference on Parallel and Distributed Processing and Applications
| Hi-index | 0.00 |
Autonomic systems require continuous self-monitoring to ensure correct operation. Available monitoring data exists in a variety of formats, including log files, performance counters, traces, and state and configuration parameters. Such heterogeneity, together with the extremely large volume of data that could be collected, makes analysis very complex. To allow for more-effective problem determination, there is a need for a comprehensive integration of management data. In addition, monitoring should be adaptive to the current perceived operation of the system. In this paper we present an architecture to meet the above goals. We leverage an open-source XML-based format for data integration and describe an approach to automatically adjust monitoring for diagnosis when anomalies are detected. We have implemented a partial prototype using an Eclipse-based open- source platform. We show the effectiveness of our prototype based on fault-injection experiments. We also study issues of disparity of data formats, information overload, scalability, and automated problem determination.