ABSTRACT
One of the most important requirements to be addressed by a general purpose Fault Management (FM) system is the ability to quickly identify the root cause of network errors and fix them as soon as possible. This informs the maintenance of an accurate model of the mobile network error logs for the FM task and the faults dynamic feature trend/statistical analysis of the error-log message summary for both the Base Station Subsystems and Mobile Switching Centers by comparing the relationship between the twin parameters of Frequency of Occurrence and the Mean Time to Repair each fault type. The Fault Trend Analysis shows that as the numbers of times a particular fault type occur increases the Mean Time to Repair such faults decreases, with some deviation from this trend in some cases when we classified them as high priority faults with serious impact on network performance and those that have little or no impact on network performance. Filtering and Correlation are two methods we used to simplify the separation of the principal alarms and redundant alarms from their side effect on network performance. These algorithms are also presented: Bayesian Inference Technique and Intelligent Management Information Base (I-MIB) MATLAB Simulation, which enabled us to systematically combine the qualitative and quantitative aspects of the Bayesian model for network fault management analysis and to reduce total computational complexity by providing a database of software alarm parameters respectively. These resulted in a Global Bayesian Network which helps to represent causal chains, i.e. links between cause/effect relationships to provide the evidence of past events and show the most likely future causes and their symptoms by computing Conditional Probabilities of each Symptoms given a particular cause i.e. using the expression P(S= Si,j / C= Ck ) to populate the Conditional Probability Table. This is critical and useful for effective Global Systems for Mobile Communications/General Packet Radio Service (GSM/GPRS) Fault Management Systems and could be deployed by Service providers for quick and proper diagnosis/troubleshooting of network event/alarm error-logs to help reduce the total downtime and improved quality of service.