Cluster computing has become an essential issue for designing immersive visualization systems. This paradigm employs scalable clusters of commodity computers with much lower costs than would be possible with the high-end, shared memory computers that have been traditionally used for virtual reality purposes. This change in the design of virtual reality systems has caused some development environments oriented towards shared memory computing to require modifications to their internal architectures in order to support cluster computing. This is the case of VR Juggler, which is considered one of the most important virtual reality application development frameworks based on open source code. This paper not only describes in detail the mechanisms based on cluster computing included in the internal design of VR Juggler, but also proposes a new global performance evaluation methodology. The goal of this methodology is to test the Preprint submitted to Elsevier Science 10 October 2007