Model-driven Instrumentation with Kieker and Palladio to Forecast Dynamic Applications
Reiner Jung, Robert Heinrich, Eric Schmieders · 2013
Abstract: Providing applications in stipulated qualities is a challenging task in today’s cloud environments. The dynamic nature of the cloud requires special runtime models that reflect changes in the application structure and their deployment. These runtime models are used to forecast the application performance in order to carry out mitigative actions proactively. Current runtime models do not evolve with the application structure and quickly become outdated. Further, they do not support the derivation of probing information that is required to gather the data for evolving the runtime model. In this paper, we present the initial results of our research on a forecasting approach that combines Kieker and Palladio in order to forecast the application performance based on a dynamic runtime model. To be specific, we present two instrumentation languages to specify Kieker monitoring probes based on structural information of the application specified in Palladio component models. Moreover, we sketch a concept to forward the monitored data to our PCM-based runtime model. This will empower Palladio to carry out performance forecasts of applications deployed in dynamic environments, which is to be tackled in future research steps. 1