Management and processing of network performance information

Intrusive monitoring systems monitor the performance of data communication networks by transmitting and receiving test packets on the network being monitored. Even relatively small periods of monitoring can generate significantly large amounts of data. Primitive network performance data are details of test packets that are transmitted and received over the network under test. Network performance information is then derived by significantly processing the primitive performance data. This information may need to be correlated with information regarding the configuration and status of various network elements and the test stations. This thesis suggests that efficient processing of the collected data may be achieved by reusing and recycling the derived information in the data warehouses and information systems. This can be accomplished by pre-processing the primitive performance data to generate Intermediate Information. In addition to being able to efficiently fulfil multiple information requirements, different Intermediate Information elements at finer levels of granularity may be recycled to generate Intermediate Information elements at coarser levels of granularity. The application of these concepts in processing packet delay information from the primitive performance data has been studied. Different Intermediate Information structures possess different characteristics. Information systems can exploit these characteristics to efficiently re-cycle elements of these structures to derive the required information elements. Information systems can also dynamically select appropriate Intermediate Information structures on the basis of queries posted to the information system as well as the number of suitable Intermediate Information elements available to efficiently answer these queries. Packet loss and duplication summaries derived for different analysis windows also provide information regarding the network performance characteristics. Due to their additive nature, suitable finer granularity packet loss and duplication summaries can be added to provide coarser granularity packet loss and duplication summaries