Stochastic reo: a case study

Abstract

QoS analysis of coordinated distributed autonomous services is currently of interest in the area of service-oriented computing and calls for new technologies and supporting tools. In previous work, the first three authors have proposed a compositional automata model to provide semantics for stochastic Reo, a channel based coordination language that supports the specification of QoS values (such as request arrivals or processing rates). Furthermore, translations from this automata model into stochastic models, such as continuous-time Markov chains (CTMCs) and interactive Markov chains (IMCs) have also been presented. Based on those results, we describe in this paper a case study of QoS analysis. We analyze a certain instance of the ASK system, an industrial software system for connecting people offering professional services to clients requiring those services. We develop a model of the ASK system using stochastic Reo. The distributions used in this model were obtained by applying statistical analysis techniques on the raw values that we obtained from the real logs of an actual running ASK system. These distributions are used for the derived CTMC model for the ASK system to analyze and to improve the performance of the system, under the assumption that the distributions are exponentially distributed. In practice, this is not always the case. Thus, we also carry out a simulation-based analysis by a Reo simulator that can deal with non-exponential distributions. Compared to the analysis on the derived CTMC model, the simulation is approximation-based analysis, but it reveals valuable insight in the behavior of the system. The outcome of both analyses helps both the developers and the installations of the ASK system to improve the performance of the system.