Modeling and simulation-driven engineering based on discrete-event systems specifications

MATIAS BONAVENTURA; DANIEL FOGUELMAN; CASTRO, RODRIGO

COMPUTING IN SCIENCE & ENGINEERING

IEEE COMPUTER SOC

Lugar: Los Alamitos, CA, USA; Año: 2016 vol. 18 p. 70 - 83

1521-9615

We present an iterative and incremental development methodology for simulation models in network engineering projects. Driven by the DEVS (Discrete Event Systems Specification) formal framework for modeling and simulation we assist network design, test, analysis and optimization processes.A practical application of the methodology is presented for a case study in the ATLAS particle physics detector, the largest scientific experiment built by man where scientists around the globe search for answers about the origins of the universe. The ATLAS data network conveys real-time information produced by physics detectors as beams of particles collide. The produced sub-atomic evidences must be filtered and recorded for further offline scrutiny.Due to the criticality of the transported data, networks and applications undergo careful engineering processes with stringent quality of service requirements. A tight project schedule imposes time pressure on design decisions, while rapid technology evolution widens the palette of network design options. Finally, due to the large scale of the project, networks and systems are available for tuning and testing only sporadically.By adopting the DEVS M&S formal framework in combination with software engineering best practices we develop network simulation models side by side with enhanced modeling capabilities and boosted simulation performance for our tools in a robust yet flexible way.We thus maximize the team?s capabilities to hypothesize and exercise candidate network design options while the real system is not available, narrowing the space of tests worth running on the real system during the scarce windows of opportunity.