Simulation is a well-established technique for analyzing systems in many domains. To analyze such systems, they are modeled at a specific level of abstraction, e.g., as a software architecture model, sequence diagram, or system-level design model. Multi-level simulation allows systems analysis based on multiple models on different levels of abstraction in a single simulation. This allows to analyze different parts of the system or the same parts in different time frames on different levels of abstraction. Therefore, the trade-off between analysis runtime and result accuracy can be adapted during the analysis. The tools used for this are traditionally monolithic with fixed levels of abstraction. We present new ideas regarding a framework for a multi-level simulation of hardware and software modeled on different levels of abstraction to evaluate quality aspects of systems, especially performance. It combines multiple independent simulators in a modular way to form an adaptable and extensible multi-level simulation that minimizes the dependencies between simulators on the different levels of abstraction.