Building a high performance computing clusters to use in computing course applications


AYDIN S., BAY Ö. F.

Procedia Social and Behavioral Sciences, vol.1, no.1, 2009 (Journal Indexed in SSCI) identifier identifier

  • Publication Type: Article / Article
  • Volume: 1 Issue: 1
  • Publication Date: 2009
  • Doi Number: 10.1016/j.sbspro.2009.01.420
  • Title of Journal : Procedia Social and Behavioral Sciences

Abstract

A cluster is a collection of independent and cheap machines, used together as a supercomputer to provide a solution. In this study, a pc cluster consisting of one master and 10 slave nodes is built to use in a cluster computing course for undergraduate and graduate students. High Performance Computing (HPC) clusters implemented by different methods are also presented here, highlighting problems, advantages and disadvantages of the methods used. Performance and ease of use of these methods are investigated in terms of their capability to solve the problems as easy as possible. Although cluster building kits like Rocks and OSCAR are easy to establish, it makes difficult to deal with problems emerging during implementation and management. It is found that HPC clusters formed by manual set-ups on Linux operating system are more useful for solving problems in more complex and flexible applications.