The deployment of modern containerized architec- tures critically depends on container orchestration tools. The performance, the robustness, and the cost of an application will hinge on the use and configuration of container orchestration tools. The reason is clear: Container orchestration tools manage scaling, health checking, failure recovery, as well as the distribution of resources between the running containers. However, it is a nontrivial and likely expensive task to evaluate configurations or even just to decide at design time already on the suitability of container orchestration as the deployment technique. These problems are exacerbated when the data required to complete such tasks are obtained through experimental analysis. Therefore, deployment of containerized architectures can be improved through modeling and simulation, as lightweight alternatives for informed decision-making in container architectures. In this paper, we present a novel approach to modeling and simulating dynamic container architectures. Our approach improves design-time decision processes and as well, the evaluation of different configurations for an existing application. Developed as an extension of the Palladio-Component-Model, our approach implements all essential mechanisms of container orchestration tools (e.g., autoscaling, container allocation scheduling, configura- tion of requests and limits). We qualitatively evaluate the approach on two example applications and load scenarios. Our experimental results demonstrate how our approach closely simulates dynamic containerized architectures while at the same time, providing lightweight comparison between different configurations and traditional deployment techniques.