Does Outdoor Computing could be part of Exascale ?

Does Outdoor Computing could be part of Exascale ?

Exascale computer is a system which is be able to execute 1 Billion of Billion computing operations per second. This is roughly the total computing power of 50 Millions laptop into a single computer. Such computing resources are used by scientists to address major challenges faced by humanity, like climate changes prediction, new energy design, transportation improvements, safety etc … Such level of computing ressources won’t be available before 2018 according to the US department of Energy, and the setup is still facing massive challenges, including system cooling, interconnect and reliability.

ASCI program is currently sponsoring US companies to develop such supercomputer into a max power request of 20MW for various physical reason, one including building cost and capability to exhaust the heat generated.

Global cost of such system migth rise up to 300 M$ including building cost, while Outdoor Computing might lower this bill by about 30%. Saved money might be re-allocated to computing ressources bringing faster systems to scientist and better return on investment, this is why we are believing that Outdoor Computing will be part of Exascale path, and that Horizon will promote such technologies for this purpose within the next couple of months starting at next OCP Summit in San Jose, and will help RuggedPOD community to progress.

We started this by providing access to our FloTherm simulation infrastructure.

You can help us to accelerate this development through our Indiegogo campaign

ps: Photo Courtesy of Oak Ridge National Laboratory, U.S. Dept. of Energy. Titan is a 27 TFlops peak Supercomputer, requires 8.2 MW of power, 404 m2 and cost 97 M$ without building. Exascale is about x40 the total power of Titan.

facebooktwittergoogle_plusredditpinterestlinkedinmail

2 Comments

  1. Here is a link to an unconventional but easier approach to exascale
    http://www.ccs3.lanl.gov/~pakin/papers/salishan2012.pdf

    > We started this by providing access to our FloTherm simulation
    > infrastructure.

    What is the bottleneck to run a simulation for 1PF/s on FloTherm?
    Does it run on a cluster? of windows HPC?

    Reply
  2. I love the guy from Los Alamos who wrote the talk you pointed us !. Currently Flotherm runs only on an SMP system, I believe the parallel version is written in poor bad OpenMP. The key challenge is more memory as a field of RuggedPOD is more a couple of millions cells rather than a few hundreds

    Reply

Submit a Comment

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

WP-Backgrounds by InoPlugs Web Design and Juwelier Schönmann