A Symptom Extraction and Classification Method for Self-Management

Marcelo Perazolo
2005 Latin American Network Operations and Management Symposium  
IT professionals often seek ways to effectively detect and prevent problems associated to the resources they manage. Their solution so far has been to rely on past experiences and established processes. Novel industry initiatives, like the Autonomic Computing paradigm provide concepts such as Events and Symptoms to facilitate automatic processing and consequently identification and treatment of problems and incidents. But the lack of an established method for identifying and classifying such
more » ... ptoms still presents itself as a complex problem. Answers would be more readily available if processes and practices were to merge into information resources comprising a generic taxonomy of symptoms and their associated knowledge types (situations and recommendations). The purpose of this study is to determine whether the essence of IT problem determination and prediction can be captured by a limited number of generic symptom types. Symptoms with nearly identical structures, such as "application connection error" and "server unreachable error" are to be classified in generic types, e.g. "network condition X". The implication of having a generic taxonomy is directly related to how easily situations can be automated and how knowledge can be identified. In this study we analyze common solutions deployed by IT shops around the world when they integrate their infrastructure with existing event management and problem determination products. The end results of this analysis and a proposal for a generic taxonomy of symptoms for IT problem determination and prediction is presented.
dblp:conf/lanoms/Perazolo05 fatcat:ke47s6devje2hkbfusmjsz3dzy