Statistical Model Checking for Product Lines
Abstract
We report on the suitability of statistical model checking for
the analysis of quantitative properties of product line models by an extended
treatment of earlier work by the authors. The type of analysis that
can be performed includes the likelihood of specific product behaviour,
the expected average cost of products (in terms of the attributes of the
products’ features) and the probability of features to be (un)installed at
runtime. The product lines must be modelled in QFLan, which extends
the probabilistic feature-oriented language PFLan with novel quantitative
constraints among features and on behaviour and with advanced
feature installation options. QFLan is a rich process-algebraic specifi-
cation language whose operational behaviour interacts with a store of
constraints, neatly separating product configuration from product behaviour.
The resulting probabilistic configurations and probabilistic behaviour
converge in a discrete-time Markov chain semantics, enabling
the analysis of quantitative properties. Technically, a Maude implementation
of QFLan, integrated with Microsoft’s SMT constraint solver Z3,
is combined with the distributed statistical model checker MultiVeStA,
developed by one of the authors. We illustrate the feasibility of our framework
by applying it to a case study of a product line of bikes.
Domains
Computer Science [cs]
Origin : Files produced by the author(s)
Loading...