Software visualization supports program comprehension by using visual metaphors that reduce textual complexity while helping developers mentally model the software. A key problem in software visualization research remains the adequate presentation of different amounts of data, that is, to achieve the desirable capability of visual scalability also for complex software systems. We present two approaches, namely semantic zoom and mini-maps, to address the challenge of visual scalability in 3D software cities. The approaches are made concrete in our web-based visualization tool, ExplorViz, which employs the city metaphor. For semantic zoom, the details and representation of elements in the software landscape are adjusted based on the virtual camera's distance from visual objects. For the mini-map, a two-dimensional top-view projection shows the essential information and structure of the software landscape in the user's proximit. Our approaches are evaluated in two different user studies. The studies indicate that both mini-maps and semantic zoom mechanisms are valuable additions to 3D software city visualizations.