Heterogeneous hardware/software systems that include many components with different characteristics offer great potential for high performance and energy-efficient computing. To exploit this potential, adaptive allocation and scheduling algorithms are needed for selecting software variants and mapping them to processing elements that attempt to achieve a good balance between resource-awareness and performance. The evaluation is typically carried out using simulation techniques. However, the space spanned by the possible combinations of hardware/software variants and management strategies is huge, which makes it nearly impossible to find an optimum using simulation-based methods. The purpose of the paper is to illustrate the general feasibility of an alternative approach using probabilistic model checking for families of systems that are obtained by varying, e.g., the hardware-software combinations or the resource management strategies. More precisely, we consider heterogeneous multi-processor systems based on tiled architectures and provide a tool chain that yields a flexible and comfortable way to specify families of concrete systems and to analyze them using the probabilistic model checker PRISM and ProFeat as a front end. We illustrate how the family-based approach can be used to analyze the potential of heterogeneous hardware elements, software variants and adaptive resource management and scheduling strategies by applying our framework to a simplified model of the multi-processor Tomahawk platform that has been designed for integrating heterogeneous devices.