Statistical Model Checking for Composite Actor Systems

Abstract

Abstract. In this paper we propose the so-called composite actor model for specifying composed entities such as the Internet. This model extends the actor model of concurrent computation so that it follows the "Reflective Russian Dolls" pattern and supports an arbitrary hierarchical composition of entities. To enable statistical model checking we introduce a new scheduling approach for composite actor models which guarantees the absence of unquantified nondeterminism. The underlying executable specification formalism we use is the rewriting logic-based semantic framework Maude, its probabilistic extension PMaude, and the statistical model checker PVeStA. We formalize a model transformation which-given certain formal requirements-generates a scheduled specification. We prove the correctness of the scheduling approach and the soundness of the transformation by introducing the notions of strong zero-time rule confluence and time-passing bisimulation and by showing that the transformation is a time-passing bisimulation for strongly zero-time rule confluent composite actor specifications.

Exploring Design Alternatives for RAMP Transactions Through Statistical Model Checking

“…Existing statistical verification techniques assume that the system is purely probabilistic. Using the methodology in [3,11] we can eliminate nondeterminism in the choice of firing rules. We then use PVeStA [4], a parallelization of the tool VeStA [23], to statistically model check purely probabilistic systems against properties expressed by QuaTEx probabilistic temporal logic [3].…”

“…The key idea to address both of these issues, following [11], is to probabilistically assign to each message a delay. The point regarding issue (ii) is that if: (a) each rewrite rule is triggered by the arrival of a message, and (b) the delay is sampled probabilistically from a dense/continuous time interval, then the probability that two messages have the same delay is 0, and hence no two actions could happen at the same time, eliminating nondeterminism.…”

Quantitative Analysis of Consistency in NoSQL Key-Value Stores

“…Existing statistical verification techniques assume that the system model is purely probabilistic. Using the methodology in [1,19] we can eliminate non-determinism in the choice of firing rules. We then use PVeStA [3], an extension and parallelization of the tool VeStA [35], to statistically model check purely probabilistic systems against properties expressed by QuaTEx probabilistic temporal logic [1].…”

“…Likewise, all nondeterministic rules in our previous model can be transformed to purely probabilistic rewrite rules. Furthermore, as explained in [1,19], the use of continuous time and the actor-like nature of the specification ensure that at most one probabilistic rule will be enabled at each time instant, thus eliminating any remaining nondeterminism from the firing of rules.…”

A White Box Perspective on Behavioural Adaptation

“…Note that, while the tower of reflection relies on a white-box architecture, the Russian Dolls approach can deal equally well with black-box components, because wrapped configurations can be managed by message passing. RRD has been further exploited for modeling policy-based coordination [73], for the design of PAGODA, a modular architecture for specifying autonomous systems [74], in the composite actors used in [32], and, by ourselves, in the design and analysis of self-assembly strategies for robot swarms [21].…”