MRI: Acquisition of a Linux Cluster Super Computer for Meteorological Modeling and Computer Science Research and Education

EIA 02-16661 苏格兰人拉夫金 圣何塞州立大学题目：MRI/RUI/MII：购置一台Linux集群超级计算机，用于气象建模和 计算机科学研究和教育项目建议：该建议来自RUI/MII机构，建立基于Linux的并行计算集群，旨在模拟气象现象并研究和教授并行处理器的使用。该基础设施将用于支持正在进行的使用火星区域大气模拟系统（MRAMS）的火星大气数值模拟和并行分布式硬件和软件的计算机科学研究。第一个组成部分将有利于以下当前正在进行的研究：火星大气的大涡模拟（LESs），以提供对高度不稳定对流边界层的结构和动力学的深入了解（包括沙尘暴的动力学）过去和拟议中的火星着陆点的中尺度模拟局部和区域沙尘暴的模拟，风力产生的地貌（辅助卫星观测）地形水冰和二氧化碳云的模拟。第二部分将研究和/或探索MRAMS代码的软件优化（在硬件平台上调整它以获得最大性能）和用于高性能计算的替代硬件架构该集群将作为中心研究平台，用于研究处理器到处理器的通信开销、对称多处理器的内存争用、缓存一致性和分布式处理。此外，新的处理器，内存和网络拓扑结构的性能将进行评估。

EIA 02-16661 Rafkin, Scot Chun, Robert San Jose State University Title: MRI/RUI/MII: Acquisition of a Linux Cluster Super Computer for Meteorological Modeling and Computer Science Research and Education Project Proposed:This proposal from an RUI/MII institution, establishing a Linux-based parallel computing cluster, aims to simulate meteorological phenomena and study and teach the use of parallel processors. The infrastructure will be used in support of ongoingNumerical modeling of the atmosphere of Mars using the Mars Regional Atmospheric Modeling System (MRAMS) andComputer Science Research in the parallel and distributed hardware and software.The first component will benefit the following current research underway:Large eddy simulations (LESs) of the Martian atmosphere to provide insight into the structure and dynamics of the highly unstable convective boundary layer (including the dynamics of dusts devils)Mesoscale simulations of past and proposed Mars landing sitesSimulation of local and regional dust storms, and wind-produced landforms (aids satellite observations)Simulations of orographic water ice and carbon dioxide clouds.The second component will investigate and/or exploreSoftware optimization of the MRAMS code (tune it for maximum performance on hardware platform) andAlternative hardware architectures for high-performance computingThe cluster will serve as a central research platform to investigate processor-to-processor communication overhead, memory contention of symmetric multiprocessors, cache coherence, and distributed processing. Moreover, performance of novel processor, memory, and network topologies will be evaluated.