Predicting Application Performance in the Cloud

Abstract

<p>Despite the exceptional prominence of the cloud computing, the customers are</p><p>lack of direct sense to select the cloud that delivers the best performance,</p><p>due to the performance heterogeneity of each cloud provider. Existing solutions</p><p>either migrate the application to each cloud and evaluate the performance</p><p>individually, or benchmark each cloud along various dimensions and predict the</p><p>overall performance of the application. However, the former incurs significant</p><p>migration and configuration overhead, while the latter may suffer from coarse</p><p>prediction accuracy.</p><p>This thesis introduces two systems to address this issue. CloudProphet predicts the web</p><p>application performance by tracing and replaying the on-premise resource demand</p><p>on the cloud machines. DTRCP further predicts the performance for general</p><p>applications. In particular, it addresses the execution path divergence</p><p>manifested during replaying the on-premise resource demand. Our experiment</p><p>results show that both systems can accurately predict the application</p><p>performance.</p>