File Exchange Pick of the Week

September 5th, 2006

Distributed computing video

I spend most of my time at The MathWorks thinking about Distributed Computing. I recently did a webinar that discusses this capability and wanted to share it with you before it goes into the recorded archives. [9/14/06: the webinar is now available in the archives, so please watch it there! (Twenty five minutes in length)]

If you are interested in getting your code to run faster, or trying to work with matrices that are bigger than you can fit into the memory of one machine, this capability is for you.

Distributed computing speed-up

By Doug Hull

13:18 UTC | Posted in Uncategorized | Permalink |

You can follow any responses to this entry through the RSS 2.0 feed. You can skip to the end and leave a response. Pinging is currently not allowed.

2 Responses to “Distributed computing video”

  1. Siamak replied on September 5th, 2006 at 15:37 UTC :

    I have two stupid questions which you are very welcome to put in any of your future books like “Distributed Computing for Dummies”

    First I was wondering if the engine is lightweight enough to run on a harddisk-less node of the cluster (I mean if the node doesn’t actually have a hard disk)

    And the second one is,
    most of the times if the code is slow I just compile it with mcc and then run the compiled version, most of the times it is faster (Or at least I feel like it is faster), my question is, can I do that with a distributed code (does mcc support distributed computing toolbox) or it should be run inside the MATLAB framework?

    By the way the very last second of the video is corrupted and I also really liked that “Confidential” thing on the slides ;)

    Thanks for the tutorial I enjoyed watching it

  2. Doug replied on September 5th, 2006 at 15:50 UTC :

    Thanks for the comments:

    A1. Buying the MATLAB Distributed Computing Engine (MDCE) allows you to do an all-product install. The only requirement is that you install MATLAB, MDCE and Distributed Computing Toolbox (DCT). I doubt you will be able to fit that into a diskless environment, but you can try!

    A2. The compiler is for deployment of applications. I would not expect it to make your code faster. To use the MATLAB Distributed Computing Engine (that is the workers on the cluster) you must have a MATLAB running with the Distributed Computing toolbox. So yes, you must be in the MATLAB framework. The tasks you give the workers can include compiled code though.

    The confidential thing was a mistake, but we can call it an “Easter Egg”

    Doug

Leave a Reply

Wrap code fragments inside <pre> tags, like this:

<pre class="code">
a = magic(3);
sum(a)
</pre>

If you have a "<" character in your code, either follow it with a space or replace it with "&lt;" (including the semicolon).

About

Bob, Brett & Jiro share their favorite user-contributed submissions from the File Exchange.

Subscribe


Recent Comments

  • Zach: Hi Doug and Les, I didn’t have a lot of time to mess with this, but I did find a work-around. I plotted...
  • hamed: k
  • Les: @Zach This isn’t exactly what you are looking for but at least it puts all three parameters on the same...
  • Zach: Thanks for your suggestions Doug. I’ll give that a shot and see what happens. I’ve seen many of...
  • Doug: @Zach, I would say to use plotYYY, because that is close to what you want, but using depth as Y makes sense....
  • Doug: @Teja, I think this will work: http://www.mathworks .com/access/helpdesk /help/techdoc/ref...
  • Gify: merry christmas :) nice christmas tree! Regards, Janet Gify
  • Teja: Dear Doug Is there anyway to plot a surface from nonuniform data without meshgrid and griddata? Basically i...
  • Zach: I’m working with geophysical data, so I’d like to produce a depth profile. The y-axis would be...
  • Doug: @Ashok First, please do not use variable names that are MATLAB commands (std and mean). Second, p(j) should be...

Blog Admin

These postings are the author's and don't necessarily represent the opinions of The MathWorks.