Dynamic graph abstraction for effective software visualisation

Abstract

Graphs are used extensively in software visualisation to represent both static aspects of software structure and dynamic aspects of execution-time behaviour. However, for realistic subject software systems, there are far too many nodes and edges in the displayed graphs to be comprehensible to an end user. Further, for presentation of dynamics, continual change and redisplay of such large graphs is too demanding for conventional workstation computational resources. This paper poses the problem of 'reduction' or 'abstraction' in dynamically changing graphs representing software systems, and proposes a combination of techniques that can be used to reduce the visual complexity of a graph, without obscuring the significant information that it was meant to convey. The abstract graph can be comprehended more readily and it changes far less frequently than the full graph. As well, when the abstract graph does change, it requires far less computation for layout and redisplay. These abstraction techniques are illustrated by way of examples showing their use in systems for visualisation of object-oriented and multi-layer software systems.