Reducing problem space using Bayesian classification on semantic logs for enhanced application monitoring and management

Monitoring managing large scale applications has always been a crucial and complex task on which enormous efforts and research have been carried out towards making the process efficient, effective and automated. However, the process is still complex, lacks efficiency and effectiveness because execution workflow representation and logging (outcome from real-time execution) is rendered in a syntactic and unstructured manner. The information is quite limited and requires additional manual interpretation till date for effectively handling the process. Hence, it makes the monitoring and management process slow, cumbersome and hard. We propose our solution by semantically (highly structured, formalized and expressive) modeling of execution workflow and logs, and then using adapted Bayesian Classification based inference technique to process formalized logs to help for enhancing the process of monitoring and management by reducing the problem space. Our hybrid approach of partially using semantics to formalize log and workflow data, and adapting classification technique combines the best of both. Semantics help in providing high-level of precision, structure and expressivity to execution workflow and logs. Such kind of formalized data can be used in an effective manner to effectively interpret and process highly structured information from the generated logs during the execution by classification technique to reduce problem space during the process of monitoring and management of applications. This paper first presents a review of related approaches, then methodology towards the hybrid solution, design of our proposed solution and implementation, followed by evaluation of our proposed solution on real-life application scenario.