Rendering computer animations on a network of workstations

Rendering high-quality computer animations requires intensive computation, and therefore a large amount of time. One way to speed up this process is to devise rendering algorithms which reduce computation by taking advantage of image characteristics, such as frame or temporal coherence. Another way is to run the rendering process on specialized or enhanced hardware, such as a multiprocessor machine. This paper details our experiences in combining these techniques through a parallel rendering algorithm exploiting frame coherence run on a network of workstations acting as a single processing system. This system provides a powerful and general method for obtaining high-quality animations with a significant reduction in computation and overall processing time. We present several techniques for partitioning the data across the workstation processors and report our results for each. While all of the partitioning schemes are scalable, some handle load balancing more effectively than others