Real-time performance modeling for adaptive software systems with multi-class workload

Modern, adaptive software systems must often adjust or reconfigure their architecture in order to respond to continuous changes in their execution environment. Efficient autonomic control in such systems is highly dependent on the accuracy of their representative performance model. In this paper, we are concerned with real-time estimation of a performance model for adaptive software systems that process multiple classes of transactional workload. Based on an open queueing network model and an Extended Kalman Filter (EKF), experiments in this work show that: (1) the model parameter estimates converge to the actual value very slowly when the variation in incoming workload is very low, (2) the estimates fail to converge quickly to the new value when there is a step-change caused by adaptive reconfiguration of the actual software parameters. We therefore propose a modified EKF design in which the measurement model is augmented with a set of constraints based on past measurement values. Experiments demonstrate the effectiveness of our approach that leads to significant improvement in convergence in the two cases.