Recent developments in the areas of computer communications have enabled the deployment of a wide variety of multimedia applications. Among the various media, video is characterized by its stringent requirements in terms of processing power, storage and bandwidth. In this paper, we undertake the study of a parallel implementation of a software MPEG-2 encoder. We use a platform consisting of a cluster of workstations interconnected by an ATM switch. The use of a parallel encoder offers great flexibility as opposed to a hardware decoder, ie, flexibility in setting parameters and modifying the various stages of the encoding process. The use of a parallel system should allow us to reduce turn around times by reducing considerable the time required to encode the video with a new set of parameters. However, the effective implementation of such application over a platform without a common memory requires a thorough analysis of the best strategy to distribute the data. This issue is particularly important in video coding, due to the large volume of video data to be handled. In this study, we pay particularly attention to the study of data allocation among the processors in order to improve the overall system operation. We explore various different methods of data distribution. Results of encoding times using different number of processors are provided. Our results show that the underlying middleware plays a major role on the performance of the overall system.