LAUSR

High-performance data streaming technologies are increasingly adopted in IT companies to support the integration of heterogeneous and possibly distributed applications. Compared with the traditional message queuing middleware, a streaming platform enables the implementation of event-streaming systems (ESS) which include not only complex queues but also pipelines that transform and react to the streams of data. By analysing the centralised data streams, one can evaluate the Quality-of-Service for other systems and components that produce or consume the streams. We consider the exploitation of probabilistic model checking as a performance monitoring technique for ESS systems. Probabilistic model checking is a mature, powerful verification technique with successful application in performance analysis. However, an ESS system may contain quantitative parameters that are determined by event streams observed in a certain period of time. In this paper, we present a novel theoretical framework called QV4M (meaning “quantitative verification for monitoring”) for monitoring ESS systems, which is based on two recent methods of probabilistic model checking. QV4M assumes the parameters in a probabilistic system model as random variables and infers the statistical confidence for the probabilistic model checking output. We present an empirical evaluation of computational time and data cost for QV4M.