The use of optimal filters to track parameters of performance models

Autonomic computer systems react to changes in the system, including failures, load changes, and changed user behaviour. Autonomic control may be based on a performance model of the system and the software, which implies that the model should track changes in the system. A substantial theory of optimal tracking filters has a successful history of application to track parameters while integrating data from a variety of sources, an issue which is also relevant in performance modeling. This work applies extended Kalman filtering to track the parameters of a simple queueing network model, in response to a step change in the parameters. The response of the filter is affected by the way performance measurements are taken, and by the observability of the parameters.