U.S.-Germany Cooperative Research on the Modelling, Simulation and Performance Analysis of an SCI-based Multiprocessor System

该奖项支持来自南卫理公会大学的Dan Moldovan教授和Eric Lin教授以及两名研究生与德国慕尼黑工业大学计算机科学系的H. Richter教授等人在计算机科学和工程研究方面进行合作。他们对并行计算系统有着共同的兴趣，特别是基于“可扩展连贯接口”（SCI）的并行计算机的体系结构。他们计划一起研究、设计和评估一个由多处理器网络组成的分布式计算机系统。这个美国小组开发了一种名为“语义网络阵列处理器”（SNAP）的并行计算机，他们希望扩大这种计算机的规模。这家德国公司已经用一个多处理器网络做了类似的工作。他们的共同努力将受益于他们在不同操作系统、网络架构和并行编程环境方法方面的互补专业知识。微处理器日益复杂的程度仍然落后于科学和工业日益增长的需求。本研究解决了与本地多处理器（LAMP）的新聚合体系结构相关的可伸缩性问题。目标是以最低的长期成本和高效率满足快速发展的网络用户需求。

This award supports Professors Dan Moldovan and Eric Lin and two graduate students, all from Southern Methodist University, to collaborate in computer science and engineering research with Professor H. Richter and others of the Computer Science Department of the Technical University of Munich, Germany. They share an interest in parallel computing systems and in particular the architecture of the `Scalable Coherent Interface` (SCI) - based parallel computers. Together they plan to study, design and evaluate a distributed computer system consisting of a network of multiprocessors. The U.S. group has developed a parallel computer called `Semantic Network Array Processor` (SNAP), which they want to scale up. The German group has done similar work with a network of multiprocessors. Their joint effort will benefit from their complementary expertise in different operating systems, network architecture and approaches to parallel programming environments. The increasing sophistication of microprocessors still lags behind the escalating demand in science and industry. This research addresses scalability questions related to a new aggregation architecture for a Local-Area Multiprocessor (LAMP). The objective is to satisfy rapidly evolving network user needs at the lowest long term cost and high efficiency.