Online Performance Prediction With Architecture-Level Performance Models
Today's enterprise systems based on increasingly complex software architectures often exhibit poor performance and resource efficiency thus having high operating costs. This is due to the inability to predict at run-time the effect of changes in the system environment and adapt the system accordingly. The authors propose a new performance modeling approach that allows the prediction of performance and system resource utilization online during system operation. They use architecture-level performance models that capture the performance-relevant information of the software architecture, deployment, execution environment and workload. The models will be automatically maintained during operation.