Chapter 7. High Performance Computing case studies

 <  Day Day Up  >  

With the rapid advances in semiconductor technology, most computing tasks that used to require supercomputers can now be performed on desktop workstations. However, there are always areas and applications that challenge and push the limit of computing power. Such applications include the "grand challenge problems" of weather forecasting, nuclear testing, drug designing, automobile designing, and so on. For these challenges, High Performance Computing (HPC) clustering is often the best solution.

An HPC cluster often consists of the following components :

  • A large number of powerful computing nodes

  • A high speed network that interconnects the computing nodes

  • A high performance global file system that is accessible from all the computing nodes

  • An efficient job scheduling system to make effective use of the cluster

  • A rich set of compilers and performance libraries for various applications

In this chapter, we address these issues for an IBM pSeries cluster running Linux, and discuss the performance of some HPC benchmarks.

The chapter includes:

  • 7.1, "Hardware and software overview" on page 305, a hardware and software overview from the HPC perspective.

  • 7.2, "Myrinet" on page 305 describes Myrinet installation and set up.

  • 7.3, "Gigabit Ethernet" on page 312 contains a brief discussion of Gigabit Ethernet.

  • 7.4, "Compilers and ESSL libraries" on page 313 discusses available compilers and their installation. We also describe how to install IBM ESSL and PESSL libraries. We present case studies for frequently used industrial standard benchmarks that also serve as a test of the installed system.

 <  Day Day Up  >  

Quintero - Deploying Linux on IBM E-Server Pseries Clusters
Quintero - Deploying Linux on IBM E-Server Pseries Clusters
Year: 2003
Pages: 108 © 2008-2017.
If you may any questions please contact us: