Large software systems typically exist in many revisions. In order to analyze such systems, their evolution needs to be analyzed, too. The main challenge in this context is to cope with the large volume of data and to visualize the system and its evolution as a whole. This paper presents an approach for visualizing the evolution of large-scale systems which uses three different, tightly integrated visualizations that are 3-dimensional and/or animated and which support different analysis tasks. According to a first empirical study, all tasks are supported well by at least one visualization.

Model-driven development leads to development processes in which a large number of different versions of models are produced. We present FAME, a tool environment which enables fine-grained analysis of the version history of a model. The tool is generic in the sense that it can work with various model types including UML and domain-specific languages.

In model-driven engineering developers work mainly or only with models, which exist in many versions. This paper presents an approach to trace single model elements or groups of elements within a version history of a model. It also offers analysis capabilities such as detection of logical coupling between model elements. The approach uses a differencing algorithm blown as SiDiff to identify similar elements in different versions of a model. SiDiff is highly configurable and thus our tracing approach can be adapted to all diagram types of the UML and to a large set of domain specific languages. The approach has been implemented as an Eclipse plug-in that visualizes all relevant information about the traces and it allows developers to interactively explore details. It has been evaluated by several groups of test persons; they considered most of the functions of the tool to be very useful.

Modern software engineering practices lead to large models which exist in many versions. Version management systems should offer a service to compare, and possibly merge, these models. The computation of a difference between large models is a big challenge; current algorithms are too inefficient here. We present a new technique for computing differences between models. In practical tests, this technique has been an order of magnitude faster than currently known algorithms. The main idea is to use a high-dimensional search tree for efficiently finding similar model elements. Individual elements are mapped onto a vector of numerical values using a collection of metrics for models and a numerical representation of the names which occur in a model.