Temporal Logic Framework for Performance Analysis of Architectures of Systems
This paper presents a formal mathematical framework for performance analysis (in terms of success of given tasks) of complex systems, ATLAS. This method interestingly combines temporal aspects (for the description of the complex system) and probabilities
- PDF / 1,296,180 Bytes
- 16 Pages / 439.37 x 666.142 pts Page_size
- 75 Downloads / 194 Views
bstract. This paper presents a formal mathematical framework for performance analysis (in terms of success of given tasks) of complex systems, ATLAS. This method interestingly combines temporal aspects (for the description of the complex system) and probabilities (to represent performance). The system’s task to be evaluated is described using a temporal language, the ATLAS language: the architecture of the task is decomposed into elementary functionalities and temporal operators specify their arrangement. Starting with the success probabilities of the elementary functionalities, it is then possible to compute the overall success probability of the task using mathematical formulae which are proven in this paper. The method is illustrated with a deorbitation task for a retired satellite called ENVISAT. Keywords: Probabilistic performance analysis tems · Temporal logic
1
·
Time-dependant sys-
Introduction
To keep up with the complexification of systems, novel performance analysis and evaluation methods have to be developed to validate new designs. In this context, architecture models of complex systems may be used to assess dynamic system performances with regard to the time necessary for the desired task to be fulfilled. The work presented here provides a generic formal framework and a tool designed for such performance analysis, called ATLAS (Analysis by Temporal Logic of Architectures of Systems). The proposed approach interestingly combines temporal and probabilistic aspects by computing the success probability of the complex system’s global task at a given instant in time and with respect to the beginning of the task. The task itself is described temporally. It is assumed that the system’s achievements may be organised as a hierarchy of functionalities: at the top, the global functionality represents the general expected behaviour of the system, i.e. its task. This global functionality may generally be split into simpler sub-functionalities, and this recursively, until reaching an elementary functionality associated to an identifiable component of the system. The success probabilities of these elementary c Springer International Publishing Switzerland 2016 S. Rayadurgam and O. Tkachuk (Eds.): NFM 2016, LNCS 9690, pp. 3–18, 2016. DOI: 10.1007/978-3-319-40648-0 1
4
A. Piel et al.
functionalities are supposed to be known. This system architecture is described by a temporal language, the ATLAS language, which allows expressing temporal constraints between the realisations of each functionality and is derived from Allen’s interval logic [1]. According to this architecture with the associated underlying temporal constraints, the global performance of the system may be computed from the individual elementary functionalities. The aim of this approach is to avoid extensive simulations and Monte-Carlo methods which are very costly in computing time. With this respect, the benefit of this method is two-fold. First of all, since the elementary functionalities are of smaller scale, if Monte-Carlo methods are necessary to
Data Loading...