Towards container-based simulation: a Kubernetes-native, container-based simulation environment

Common simulation studies, simulation-based optimization, data farming, and machine learning benefit strongly from large numbers of simulation runs. The required computational resources frequently exceed the capabilities of traditional office environments, motivating the transition toward distributed computing infrastructures. To address this demand, we present a concept for a Kubernetes-native, Container-Based Simulation Environment (CBSE) designed to facilitate scalable and efficient execution of simulation workloads. We provide a historical overview of the evolution of parallel replications, highlighting the transition toward container-based approaches. Building on this foundation, we derive requirements for a generalized simulation platform, as well as conditions for the integration and containerization of simulation software. The proposed architecture extends the Kubernetes API with simulation-specific custom resources, orchestrated through the Experiment Operator and coordinated by the Scenario Manager. These components are intended to streamline experiment definition, automate scenario execution, and enable robust scheduling mechanisms. By leveraging Kubernetes as its underlying infrastructure, the approach is intended to support deployment across a range of environments, from local clusters to large-scale distributed systems. Although our reference implementation is still under development, this paper presents and discusses the underlying concepts and our plans for the future.