Maintaining dynamic geometric objects on parallel processors

Parallelizing dynamic scientific applications involves solving the dynamic load balancing problem. The balancing should take the communication requirements of the application into account. Especially in the field of computer graphics, many problems deal with objects in k-dimensional space with very special communication patterns. We describe a load balancing strategy, based on a geometric clustering of objects, with provable good performance, that is well suited to applications with interdependencies between neighbouring objects or between objects that have at least one coordinate in common.