Modelling dynamic e-business applications using historical performance data

Distributed e-business application platforms such as the web services framework will require sophisticated workload management infrastructures for the matching of client requests to the most appropriate resources. This paper introduces a novel dynamic predictive framework that uses both historical data and analytical performance modelling to extrapolate a requestís predicted performance under varying workloads and conditions on different systems. Such a framework can facilitate a workload manager in maximising the extent to which service level agreements can be met. Initial results are presented analysing how the response times of an e-business benchmark can be predicted at different system loads and on different application servers