For most embedded safety-critical systems not only the functional correctness is of importance, but they must provide their services also in a timely manner. Therefore, it is important to have rigorous analysis techniques for determining timing properties of such systems. The ever increasing complexity of such real-time systems calls for compositional analysis techniques, where timing properties of local systems are composed to infer timing properties of the overall system. In analytical timing analysis approaches the dynamic timing behavior of a system is characterized by mathematical formulas abstracting from the state-dependent behavior of the system. While these approaches scale well and also support compositional reasoning, the results often exhibit large over-approximations. Our approach for compositional timing analysis is based on ω-regular languages, which can be employed in automata-based model-checking frameworks. To tackle the scalability problem due to state-space explosion, we present a technique to abstract an application by means of its resource demands. The technique allows to carry out an analysis independently for each application that shall be deployed on the same platform using its granted resource supply. Integration of the applications on the platform can then be analyzed based on the different resource supplies without considering details of the applications.