Parallel Programming on Workstation Clusters

工作站集群上的并行编程

• 批准号: 9950309
• 负责人: Anthony Wilkinson
• 金额: $ 10.02万
• 依托单位: University of North Carolina at Charlotte
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 1999
• 资助国家: 美国
• 起止时间: 1999-06-01 至 2002-05-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9950309&HistoricalAwards=false
• 关键词: Parallel; Programming; Workstation; Clusters

This project impacts upon how parallel programming is taught using workstation clusters, the most common computing platform for undergraduate institutions. It utilizes the recent introduction of inexpensive multiprocessor workstations and clusters of such multiprocessor systems. Conventionally, parallel programming on workstation clusters has been done using the message-passing paradigm. However, the message-passing paradigm has serious disadvantages from a programming perspective, including requiring message passing to be specified explicitly in the program, and having to take into account very significant communication delays. It is widely accepted that a preferred programming paradigm is the shared memory paradigm. This is especially so for undergraduate students who have already been taught and understand sequential programming. This project explores the use of shared memory programming on workstation clusters, concentrating upon thread-based parallel programming. New educational materials for undergraduate teaching, including necessary software tools for distributed shared memory, are being developed to support this approach. This work is intended to make a change from purely message passing programming on workstation clusters to shared memory and hybrid shared memory/message-passing programming.In addition to developing new courses for the local students, the courses are being offered on the NC-REN, a teleclass network for several North Carolina universities. Another test site for the course will be a local historically black university. A professor from that university will be involved in course development and training to enable him to teach and assess the course at his own institution. If the proof-of-concept of this approach is successful, teaching materials will be published in a textbook and made available through a web site.

这个项目影响如何使用工作站集群，本科院校最常见的计算平台并行编程教学。 它利用了最近推出的廉价的多处理器工作站和集群等多处理器系统。传统上，工作站集群上的并行编程是使用消息传递范式完成的。然而，从编程的角度来看，消息传递范式有严重的缺点，包括需要在程序中显式地指定消息传递，并且必须考虑非常显著的通信延迟。广泛接受的是，优选的编程范例是共享存储器范例。 对于已经接受过教学并了解顺序编程的本科生来说尤其如此。这个项目探讨了在工作站集群上使用共享内存编程，专注于基于线程的并行编程。 正在开发新的本科教学教材，包括分布式共享内存的必要软件工具，以支持这种方法。 这项工作的目的是作出改变，从纯粹的消息传递编程工作站集群共享内存和混合共享内存/消息传递programming.In除了开发新的课程，为当地学生，这些课程正在提供的NC-REN，一个teleclass网络的几个北卡罗来纳州大学。 该课程的另一个考点将是当地一所历史悠久的黑人大学。 该大学的一名教授将参与课程编制和培训，使他能够在自己的机构教授和评估课程。 如果这一方法的概念验证成功，将在教科书中出版教材，并通过网站提供。