Have your browser do some work for a change! Help some GUTS-XL/Challenge students

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

Have your browser do some work for a change! Help some GUTS-XL/Challenge students

Stephen Guerin
Please help out Max and Harsha test their distributed computing  
project. Point your browser to http://mds.g-r-c.com/ and leave it  
running while you're not using your machine just for the next day or  
so. More info below from their mentor, Nick Bennett.

-Stephen


> From: Nick Bennett <[hidden email]>
> Date: March 31, 2009 7:56:18 PM MDT
> To: Joshua Thorp <[hidden email]>,  Irene Lee <[hidden email]
> >, Paige Prescott <[hidden email]>,  talaya white <[hidden email]
> >, Steve Guerin <[hidden email]>
> Cc: Meestuh Boing <[hidden email]>, [hidden email],  
> robert bond <[hidden email]>
> Subject: Help some GUTS-XL/Challenge students, and have your browser  
> do some work for a change!
>
> Folks,
>
> Max and Harsha, two of our GUTS-XL students, are doing some almost-
> final crunching on their project - which is not only their project  
> for GUTS-XL, but also their Challenge project, as well as Max's  
> project for his mentorship at Monte del Sol this year.
>
> The project is on distributed computing via applets in web pages.  
> This approach would allow a much wider range of machines than usual  
> to participate as computing resources in solving certain kinds of  
> computationally intensive problems. Also, since the applets in  this  
> case are communicating with the task distribution server via XML-RPC  
> over HTTP, no special firewall policy or router port configuration  
> is required to allow participation. Basically, if a user can browse  
> to the web page hosting the applet, and if the user has Java 5 (1.5)  
> or higher installed, it's virtually certain that the user's machine  
> can participate - no matter where in the world the user is, and no  
> matter how the user is connected to the Internet.
>
> Variations on this kind of distributed computing have been used in  
> SETI@HOME, GIMPS (Great Internet Search for Mersenne Primes),  
> cracking some of the final remaining Enigma cyphertext messages from  
> WWII, and other projects. However, virtually all of these have been  
> more restrictive in their approach, generally requiring the  
> installation of special software on the user's machine (often  
> restricted to a subset of the operating systems and platforms in  
> wide use), and sometimes special configuration of any firewalls the  
> user might be behind. Max & Harsha's project is thus an attempt to  
> demonstrate the opportunity to extend such projects to a much more  
> heterogeneous world of platforms and networks.
>
> For the proof of concept, they've chosen the search for Mersenne  
> primes - partly because this is an interesting topic in and of  
> itself, but also because it's a well-understood problem, with  
> interesting mathematical and algorithmic aspects, and with an  
> existing distributed computing project (GIMPS) that can be used as a  
> point of comparison and contrast.
>
> So my suggestion - to those of you who feel like letting your  
> computer participate as a resource for some portion of the next 18  
> hours (as Max and Harsha gather the last round of data for their  
> report), and if you can spare the cycles - is to point your  
> JavaScript- and Java-enabled (v5/1.5 or higher) browser to http://mds.g-r-c.com/ 
>  - that's it, that's all it takes. All you'll see, as your computer  
> participates in the process, is a status message that changes from  
> time to time, and a task history that gets updated every few minutes.
>
> At this point in the search, testing each candidate Mersenne numbers  
> for primality is a slow process, since we're dealing with very large  
> numbers (i.e. those currently being tested have over 5,000 digits  
> when written out in base 10 form). Please note that these tests are  
> computationally intensive: if you have a single-core CPU, Java will  
> pretty much try to use all of its power; so if you need to do a lot  
> of work on your computer, it would probably be less annoying for you  
> if you don't participate at this point. But the power of the  
> distributed approach to this problem is that each participant  
> resource can be working on a different Mersenne number, cutting down  
> the time to move through the search space. So, if you have the  
> cycles to spare, join in!
>
> Thanks and regards,
>
> Nick
>
>
>


============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org