Operators of cellular networks are hard pressed to provide a seamless wireless connection to their users. Due to the expanded demand not only for coverage but also for increased network capacity, the network architecture needs to be adapted and evolve beyond the classical hexagonal grid. The globally ongoing trend of urbanization leads to more and more users utilizing their wireless devices indoors or in mobile scenarios, when commuting or traveling. These scenarios pose particular challenges to implementing a suitable network in terms of propagation conditions as well as optimal base station (BS) deployment. Therefore, in this thesis, I investigate the potential network-wide average performance of wireless cellular networks particularly in high speed train (HST) environments , as well as of network deployments indoors. An investigation on network scale requires to limit the complexity of the applied system models. This is necessary in order to still obtain mathematically tractable formulations as well as to be able to perform simulations with finite duration. Since the scenarios under investigation are in themselves rather specific and also in comparison fundamentally different for some aspects, I first introduce available models for signal propagation and interferer geometry. I then justify the chosen models and introduce the different approaches for evaluating performance metrics in both scenarios. Wireless cellular networks in HST scenarios exhibit several specific aspects of technical and non-technical nature that need to be taken into account in the system design. Especially the decision between direct communication or relay aided communication has to be made. To justify the utilization of system level (SL) simulations, that include various abstractions, throughput results are compared to more detailed link level (LL) simulations, parameterizing both simulations with reference values from real-world measurements. Additionally, I present a framework for investigating the performance improvement of remote unit collaboration schemes. Signal propagation in indoor scenarios is dominated by wall blockages. Initial investigations focus on the influence of the distribution of blockage objects. Therefore, four different wall generation methods are introduced, which are parametrized such that the created spatial scenarios remain comparable. Based on these methods, expressions for the average attenuation of a link and for the signal to interference ratio (SIR) are derived. Subsequently, the effect of the BS placement is considered, by placing BSs randomly or by arranging them in a regular grid. Here, additionally the area spectral efficiency (ASE) is utilized to identify the dependency of the network capacity on the BS density.