In this paper a new approach for a metrics based software visualization is presented which supports an efficient and effective quality assessment of large object-oriented software systems. It is based on the combination of software metrics data with structure information to form a virtual information space. This information space is visualized using 3D graph structures that allow to represent in a uniform way many aspects and views on it. The layout approach for these graphs uses a generic similarity measure to calculate geometric distances between the graph nodes and a force-directed mapping into 3D space. A particular strength of the approach is that the resulting geometrical structures can be well interpreted with respect to the architecture and design quality of the analysed software. Such 3D visualizations have been used successfully in several case studies for the quality assessment of industrial C++/Java projects.