There is an interesting parallel between building Ferraris and building computers. Both require elements to make them very fast and very complex. Like computers, Ferraris have a lot of commodity components such as tires. But the way these components are put together and the resulting end system is what makes them unique. The high performance computing space is no different.

Sun's Director of Technology in the Global Systems Practice, Marc Hamiltons's Blog, meets with Hal Stern, vice president of global systems engineering, in this edition of Innovating@Sun to discuss the art of tying together commodity components to get the most compute, networking, and storage in the smallest space.

Hamilton and Stern discuss what the high-performance computing (HPC) market looks like today with a focus on how technical problems today needs more capacity than you can deliver on a single desktop. Highlights of the interview include:

  • What mechanisms in Solaris make it possible to scale with increased clustering
  • Market interest around HPC given that applications are expanding and becoming more data and network driven
  • Mitigating silent data corruption with ZFS
  • Accounting for hardware failures in your software architecture to reduce the frequency of system failures
  • Multicore chips that are increasing demands on the operating system
  • The ever-present issues of power and cooling and what new innovations address these
  • Fortress – a modern language for HPC programming

    Links:



    Show Transcript
    Marc Hamiltons's Blog
    Open MPI
    X4500 storage server
    Andy Bechtolsheim
    Fortress
    Guy Steele