Object-oriented clusters : a proposal for a new parallel processing paradigm = 目标导向群 : 关于新并行处理范例的研究

Clustering desktop computers in order to achieve cheap ’supercomputing power’ has become a popular approach for processing labor intensive tasks for example in the academic and entertainment fields during the past few years. The distributed architecture and scalability of PC clusters provide a highly suitable hardware foundation in particular for the processing of large sets of data that are easily broken into pieces and distributed for parallel computation. However, required advanced programming skills, development time as well as data and algorithms that are not easily parallelized （as typically found in real time shared virtual reality generation） are major obstacles on the pathway to flexible cluster application.In this paper, we propose a new cluster topology that takes advantage of the Java 3D data architecture to generate large virtual reality datasets. A discussion of our experimental investigation into this approach is given with special emphasis on our design perspective towards virtual reality production. The benefits of the proposed strategy are manifold and include easier and more economic development at comparatively gentle learning curves, significantly better runtime performance, and the reusability of generic software components. This can contribute to a wider range of real time VR content to be produced and hence expand the application of virtual reality beyond the boundaries of the design field.

为取得“超计算功能”，PC电脑群已成为一种很受欢迎的方法，这种方法能够处理诸如在过去几年里学术领域和娱乐领域的劳动密集型工作。而分布式的体系结构和可测量性的个人电脑为并行计算过程中易分散的大量固定数据提供了十分合适的硬件基础。然而，这一过程需要先进的编程技巧，大量的调试时间以及数据处理和计算规则并行的困难性 (实时共享虚拟现实产生的典型) 是灵活运用导向群的主要障碍。设计一个新的拓扑群，这一拓扑群利用Java 3D数据结构以产生大量虚拟现实数据结构。但是专家们对实验研究中给予特别重视的关于虚拟现实作品的设计方法产生了争议。在相对平缓的识别图表曲线的过程中，值得关注的是执行程序的运行时间以及软件类的组成部分的可利用性。这有助于建立一个广范围的实时的虚拟现实环境，以促进设计方案的提出，从而能越过设计领域，扩大虚拟现实的应用范围。