Paper Authors

Abstract

NOTE: The first page of text has been automatically extracted and included below in lieu of an abstract

Developing a Course for Hands-on High-Performance Computing

Abstract

High-performance computing (HPC) based on commodity hardware and open-source software has become the dominant paradigm for supercomputing today.1, 2 Thus a great unmet need exists for skilled students and practitioners who can design, develop, deploy, and operate HPC-based systems to support discoveries in industry and academe.

To address these needs, we have developed two courses in HPC, one for undergraduates and one for graduate students, that provides students with hands-on experience in designing, developing, and testing commodity-based supercomputing systems. In this paper, we describe a cost- effective and scalable approach that we developed for this course, which has been successfully delivered over several semesters. We describe the curricular context, pedagogical approach, and outcomes along with a detailed description of the approaches and strategies we used to develop a hands-on laboratory component that can be replicated by others seeking to develop similar courses. We believe that our results will be useful to departments and institutions interested in developing curricula to answer the increasing needs presented by HPC and cyberinfrastructure.

Introduction and Motivation

In the race for global competitiveness in technology, manufacturing, science, and engineering, computer-based design and simulation have become critical elements in producing higher quality and less-expensive products. Computer simulation and data analysis are central to this effort. Computer simulation is used by automobile manufacturers to design better products in shorter times at lower costs, to discover new and previously overlooked sources of oil and gas, and to improve industrial processes.3 Data analysis, which involves sifting through terabytes of data to discover trends and unexpected patterns, is another emerging area of computation that is helping to improve product design, science, and engineering research.

Algorithms that form the “computational engine” and heart of these analysis and simulation efforts have grown in capability and complexity as the physical and temporal scales and resolutions of the problems posed by researchers increase. The increasing computational scale and complexity have created an increased demand for greater computational capacity and storage resources. This trend in computational demand has naturally led to a significant increase in the use of HPC and supercomputing to provide an effective computational platform on which to solve these problems.4-9 In the past, supercomputing systems were designed and built exclusively from custom-engineered components, and thus they were rare and expensive. The powerful trends of increasing computational power and decreasing costs of commodity computer components driven by the home computing and gaming markets have completely changed the supercomputing industry over the past 15 years. As noted in a recent National Research Council Report10: “…over the last several years such commodity supercomputers have rapidly come to dominate the supercomputer marketplace.” The dominance of commodity-based cluster computing systems is evidenced by the growing fraction of cluster-based systems on the Top 500 list – a semiannual list of the 500 fastest supercomputers.1 In November 2000, a few years after

• Citation

• Format
  • APA
  • APA - LaTeX bibitem
  • MLA
  • MLA - LaTeX bibitem
  • Bibtex
  • EndNote - RIS

Hacker, T. (2010, June), Hands On High Performance Computing: Developing A Cluster Computing Course For Real World Supercomputing Paper presented at 2010 Annual Conference & Exposition, Louisville, Kentucky. 10.18260/1-2--16135