METHODS AND ALGORITHMS FOR ANALYSING AND PREDICTING ERRORS IN SERVER SYSTEMS.

Logging is important in monitoring and predicting server failures. The information is stored in log files, which contains (usually as plain text) information about messages, error reports, file requests and file transfers. Reasons for logging are diagnostic, audit, capturing stack traces or data. The purpose of this work is to develop a software architecture for analysis and prediction of errors in server systems. First, messages are collected from client servers and stored on the main server. Then a message parser recognizes the lexical and syntactic structure by using context-free grammar. Key components are stored in a relational database. A synthesizer is connected to a relational database. It is used to recreate the original messages when a user needs to read them. The most important part of this work is the recognition of interconnections between messages and environment with the help of artificial intelligence, which will thereafter be used to predict failures. [ABSTRACT FROM AUTHOR]