A knowledge based problem diagnosis apparatus through machine learning
Publication Date: 2016-May-19
The IP.com Prior Art Database
More and more enterprise infrastructure systems are moved into a cloud environment, which consists of hardware, network, management software, application, etc. Identifying root cause from a failure case in such a complex environment is a headache to most operation teams. This article describes an appratus to identify potential root causes automatically based on a knowledge base if an incident occurs. In this appratus, controlled chaos are generated and the symptoms for a failure ( including the errors that the user experience, and the monitoring items that are collected from monitoring compoents) are collected. The mapping between the symptoms and the root cause are mapped together and stored as a RAW pattern in a knowledge base. RAW patterns are analyzed automatically and the probability of a pattern occuring is computed and stored as a FINAL pattern. After the knowledge base is built, analyed and refined automatically, it could be used in a production environment. When an incident occurs, the symptoms are collected and described as a case. A CBR (Case Based Reasoning) algorithm is used to find the most similar pattern in the knowledge base. Therefore, the RC (Root Cause) with largest probability is sugested to the operation team for further action.