부산대학교 도서관

Many algorithms in the field of scientific visualization require high computational capacity. In general, low-end machines are not capable of delivering sufficient performance to achieve interactivity. Even worse, to generate high quality images of large data sets, the necessary time is in the order of minutes. Besides algorithmic optimizations, parallelization is necessary to quicken the computation to enable some interactivity to comprehensively understand the volumetric data. In this paper we present the implementation of two visualization algorithms using PVM. To circumvent stall conditions and to guarantee a somehow continuous time per frame, we propose different scheduling schemes. Both algorithms are tested using the different strategies in a multi-user environment and in a dedicated network of workstations.