Efficient Synthesis of Robust Models for Stochastic Systems

Software performance and reliability engineering, 
Probabilistic model synthesis, 
Multi-objective optimisation, 
Robust design

We describe a tool-supported method for the efficient synthesis of parametric continuous-time Markov chains (pCTMC) that correspond to robust designs of a system under development. The pCTMCs generated by our RObust DEsign Synthesis (RODES) method are resilient to changes in the systems operational profile, satisfy strict reliability, performance and other quality constraints, and are Pareto-optimal or nearly Pareto-optimal with respect to a set of quality optimisation criteria. By integrating sensitivity analysis at designer-specified tolerance levels and Pareto optimality, RODES produces designs that are potentially slightly suboptimal in return for less sensitivity-an acceptable trade-off in engineering practice. We demonstrate the effectiveness of our method and the efficiency of its GPU-accelerated tool support across multiple application domains by using RODES to design a producer-consumer system, a replicated file system and a workstation cluster system.

Robustness is a key characteristic of both natural and human-made  systems. Systems that cannot tolerate changes are prone to frequent failures and require regular maintenance. We describe a tool-supported method for the efficient synthesis of parametric continuous-time Markov chains (pCTMC) that correspond to robust designs of a system under development (SUD). The obtained designs are resilient to changes in the system's operational profile, satisfy strict reliability and performance constraints, and are nearly Pareto-optimal with respect to a set of quality optimisation criteria.

Our RObust DEsign Synthesis (RODES) method comprises two steps. In the first step, the SUD design space is modelled as a pCTMC with discrete and continuous parameters corresponding to alternative system architectures and to ranges of possible values for the SUD parameters, respectively. In the second step, a multi-objective optimisation technique leveraging parametric analysis of CTMCs is used to obtain a set of low-sensitivity, nearly Pareto-optimal SUD designs by fixing the discrete parameters (thus selecting specific architectures) and restricting the continuous parameters to bounded intervals that reflect the pre-specified tolerances. A sensitivity-aware Pareto dominance relation is introduced  to formally capture trade-offs between the robustness and optimality of SUD. 

To the best of our knowledge, RODES is the first solution that integrates multi-objective stochastic model synthesis and sensitivity analysis into an end-to-end, tool-supported design method. We demonstrate the applicability of the method and the efficiency of its GPU-accelerated tool support using three case studies from different  engineering  domains.

Citations (including self-citations): Google Scholar 12, Scopus 6.

Remark (not sure if this should be mentioned):

The paper integrates and extends the results published in the paper Designing robust software systems through parametric Markov chain synthesis at IEEE International Conference on Software Architecture (ICSA), 2017 (Citations (including self-citations): Google Scholar 14, Scopus 8) and in the tool paper RODES: A robust-design synthesis tool for probabilistic systems at International Conference on Quantitative Evaluation of Systems (QEST 2017) (Citations (including self-citations): Google Scholar 5, Scopus 2)

Robustness is a key characteristic of systems allowing to tolerate changes in their parameters, operational profile and environment. We describe a novel method allowing to automatically design systems that are robust (i.e. resilient to these changes), satisfy strict reliability and performance constraints, and are nearly Pareto-optimal with respect to a set of quality optimisation criteria. To effectively reason about trade-offs between the robustness and optimality of the system under development, we introduce a sensitivity-aware Pareto dominance relation. We integrate the relation  into multi-objective synthesis of parametric stochastic models. We demonstrate the applicability of our method and the efficiency of its GPU-accelerated tool support using case studies from different  engineering  domains.

Citations (including self-citations): Google Scholar 12, Scopus 6.