Multiprocessor and distributed system design: The integration of functional specification and performance analysis using Stochastic Process Algebras

Abstract

We introduce Stochastic Process Algebras as a novel approach for the structured design and analysis of both the functional behaviour and performance characteristics of parallel and distributed systems. This is achieved by integrating performance modelling and analysis into the powerful and well investigated formal description technique of process algebras. After advocating the use of stochastic process algebras as a modelling technique we recapitulate the foundations of classical process algebras. Then we present extensions of process algebras such that the requirements of performance analysis are taken into account. Examples illustrate the methodological advantages that are gained. 123 1.2 Precise and Modular Description of System Behaviour Classical queueing and loss models are the best known and most often used performance evaluation techniques. They mainly use a mixture of (almost) standardized box symbols, precisely defined stochastic parameters and colloquially formulated scheduling strategies. These semi-formal techniques have been most successfully applied to describe and analyse the behaviour of many dynamic systems. Modelling, however, the complex interdependencies within distributed and parallel processor systems we need a more precise, expressive and unambiguous description. Formal languages and particularly process algebras (two of them, CSP [Hoa85] and CCS [Mi189], are the basis of the quite remarkable programming languages OCCAM and LOTOS) are best suited and generally accepted as appropriate means. Stochastic Process Algebras areas we will elaborate during this tutorial-an extension of these classical abstract languages including random time variables and operators for the description of probabilistic behaviour. Some examples are shown below to illustrate intuitively their expressive power. (Here we assume exponentially distributed time intervals, characterized by their rates; in general these rates just have to be replaced by an appropriate parameter set.) 9 The sequential arrival of three different jobs is specified by a process Jobstream describing explicitly each arrival point before halting:

Validating Distributed Object and Component Designs

“…It would be highly desirable to avoid costly risk mitigation iterations during a development process and address the question of whether an architecture scales and performs efficiently and reliably by analytic means. The performance modelling literature includes a large body of work on stochastic process algebras, which use distribution functions with which transitions are executed [12,9,8]. It seems natural to extend the research that we presented here to performance, scalability and reliability properties of UML models that can then be expressed and analysed with stochastic process algebras.…”

Simulation and Statistical Model Checking of Logic-Based Multi-Agent System Models

“…However, extensions exist that feature these concerns forming a relatively large family of stochastic process algebras like TIPP [GHR93], PEPA [Hil96], EMPA [BG96,ACB10], stochastic π-calculus [Pri95,CM13], StoKlaim [DNKL + 07] or a unifying framework by Nicola et al [NLLM13]. Targeting the framework of SCEL, which is arguably closest to the approach presented here, the ideas of stochastic process algebras are well integrated in [L + 14], where the authors introduce a stochastically timed process calculus that is centered around predicate-based communication.…”

Towards Testing Stochastic Timed Systems

“…Thus, stochastic extensions of classical formal models, as process algebras and Petri Nets, have appeared in the literature (see e.g. [1,2,5,7,11,12,14,16,20]). As we have shown in the previous example, the main idea underlying stochastic models is that time information is incremented with some kind of probabilistic information.…”