Construction of National Grid and Development of Network Worm type Particle-In-Cell code over it

网格计算是一个源于“电网”的术语。其想法是，只要支付计算费，计算能力就会在需求和供应基础上提供。万维网服务仅限于信息服务。另一方面，网格计算将为用户提供计算能力、可视化、信息和感知等全方位的协作服务。网格计算还将包括在广域网上共享文件和应用程序的想法。介绍了我们在学术高性能网络SuperSINET上进行的基于计算网格系统的并行Partic-In Cell模拟实验。计算网格系统由SuperSINET上的三个计算机节点组成，配置了Clobus工具包和MPICH-G2。在我们的实验中，我们首先验证了并行PIC模拟代码在这个广域网格系统上的可扩展性。在对测试结果进行分析后，发现网络通信存在瓶颈，并成功地利用Globus工具包和MPICH-G2的一些特性来提高网络的计算效率。

Grid Computing is a term originated from "Power Grid". The idea is that the computing power is provided on the demand and supply base so long as the computing fee will be paid. World Wide Web services are limited to those of information. On the other hand, Grid computing will provide total cooperative services that users can benefit including computing power, visualization, information, and sensing. Grid computing will also include an idea including sharing files and applications over wide area networks. We present our parallel Partic-In Cell simulation experiment based on Computational Grid system on the academic high performance network SuperSINET. Computational Grid system consists of three computer nodes on SuperSINET, which configured with the Clobus Toolkits and MPICH-G2. In our experiment, we first conform the scalability of the parallel PIC simulation code on this wide area Grid system. After we analyze the results and find that the bottleneck point exist on the communication across the network, we successfully make use of some features in Globus toolkit and MPICH-G2 to improve the computational efficiency across the network.