Design of a Model Execution Framework: Repetitive Object-Oriented Simulation Environment (ROSE)

The ROSE framework was designed to facilitate analyses of complex systems. It completely divorcesthe model execution process from the model itself. By doing so ROSE frees the modeler to develop alibrary of standard modeling processes such as design of experiments, optimizers, parameter studies, andsensitivity studies which can then be applied to any of their available models.The ROSE framework accomplishes this by means of a well dened API and object structure. Both theAPI and object structure are presented here with enough detail to implement ROSE in any object-orientedlanguage or modeling tool. 1 Introduction Today, state of the art systems analysis methods revolve around the ability to execute a given system modelusing statistical methods and optimizers. These methods need to be repeatable and widely applicable to arange of models, regardless of what modeling environment is used. There are many different programminglanguages and as many different modeling environments currently in active use today. C++, Java, Python,Fortran, Simulink, Model Center, iSight, and NPSS are just are just a few from the available languageand modeling environment options. With so many options, there is a need for a unifying structure todene analysis processes and solution paths which can be easily applied in any language or modelingenvironment available.A second feature of almost any modern analysis methodology is the large case sets required to completethem. Optimizations and statistical methods all rely on the computational power available today to exe-cute many simulations of a given model in order to obtain the necessary information about it’s behavior.Very often, the many individual simulations are completely independent of each other and can be executedconcurrently, assuming the necessary parallel computing resources are available. In recent years, large clus-ters of inexpensive servers have provided an unprecedented level of parallel computing power. Similarly,desktop workstations now come equipped with multi-core CPU’s which can make use of parallelized pro-cesses. Any model execution framework needs to be able to seamlessly take advantage of these types ofcomputation resources as they become available.The ROSE framework was designed primarily to serve the needs of a modern systems analysis method-ology. It addresses both the issue of creating general repeatable methodologies, as well as challenges in-volved with utilizing parallel computing resources. Presented here is a structural outline of the ROSEframework that is detailed enough to implement it in the language and/or modeling environment of yourchoice, along with an example implementation written in Python.