Model-Driven Performance Prediction of Systems of Systems

International audienceSystems of Systems exhibit characteristics that pose difficulty in modelling and predicting their overall performance capabilities, including the presence of operational independence, emergent behaviour, and evolutionary development. When considering Systems of Systems within the autonomous defence systems context, these aspects become increasingly critical, as performance constraints are typically driven by hard constraints on space, weight and power. System execution modelling languages and tools permit early prediction of the performance of model-driven systems, however the focus to date has been on understanding the performance of a model rather than determining if it meets performance requirements, and only subsequently carrying out analysis to reveal the causes of any requirement violations. Such an analysis is even more difficult when applied to several systems cooperating to achieve a common goal - a System of Systems (SoS). The successful integration of systems within a SoS context has been identified as one of the most substantial challenges facing military systems development [2]. Accordingly, there is a critical need to understand the non-functional aspects of the SoS (such as quality of service, power, size, cost and scalable management of communications), and to explore how these non-functional aspects evolve under new conditions and deployment scenarios. It is crucial that we develop methodologies for modelling and understanding non-functional properties early in the development and integration cycle to better inform our understanding of the impact of emergent behaviour and evolution within the SoS. We propose an integrated approach to performance prediction of model-driven real time embedded defence systems and systems of systems [1]. Our architectural prototyping system supports a scenario-driven experimental platform for evaluating model suitability within a set of deployment and real-time performance constraints. We present an overview of our performance prediction system, demonstrating the integration of modelling, execution and performance analysis, and discuss a case study to illustrate our approach. Our work employs state-of-the-art model-driven engineering techniques to facilitate SoS performance prediction and analysis at design time, either before the SoS is built and deployed, or during its lifetime when required to evolve. Our model-driven performance prediction platform supports a scenario-driven experimental environment for evaluating a SoS within the context of a specific deployment (modelling geographical distribution) and integration constraints. The main contributions of our work are: (a) a modeling methodology that captures diverse perspectives of the performance modeling of Systems of Systems; (b) a performance analysis engine that captures metrics associated with these perspectives and (c) a case study showing the performance evaluaton of a system of systems and its evolution as a result of the performance analysis. We discuss how our approach to modelling supports the specific characteristics of an SoS, and illustrate this through a case study, based on a "Blue Ocean" scenario, demonstrating how we may obtain performance predictions within a SoS with emergent and evolutionary properties. Within the context of our environment, we define models for the individual systems within our System of Systems, defined for representative workload to predict execution costs, i.e. CPU, memory usage and network usage, within a generic situation. Our modelling environment supports the generation of executable forms of these models, which may then be executed above realistic deployment scenarios in order to obtain predictions of System of System performance