[astro] Re: boinc update

[Carlos A. Varela]

It is indeed impressive that we have more processing power than the Blue 
Gene with 1024 processors (with a single BOINC server).  If we can learn 
to use this power for non-massively parallel applications as well 
(combining BOINC and IOS), it would be a big hit!

Cheers,

- Carlos.

---------------------------------------------------------------------
Carlos A. Varela      <cvarela at cs.rpi.edu>      Ph: +1 (518) 276-6912
Associate Professor                            Fax: +1 (518) 276-4033
Worldwide Computing Laboratory                 http://wcl.cs.rpi.edu/
Department of Computer Science                      Office: Lally 308
Rensselaer Polytechnic Institute, 110 8th Street, Troy, NY 12180, USA
---------------------------------------------------------------------

On Mon, 19 Nov 2007, Boleslaw Szymanski wrote:

> this is impressive indeed and exciting!
>
>
> ---------------------------------------------------------------------------
> Boleslaw K. Szymanski
> Claire and Roland Schmitt Distinguished Professor of Computer Science
> Rensselaer Polytechnic Institute      e-mail: szymansk at cs.rpi.edu
> Lally 204, 110 Eighth Street          http://www.cs.rpi.edu/~szymansk
> Troy, NY 12180-3590, USA              tel: 518-276-2714, fax: 518-276-4033
>
>
> On Mon, 19 Nov 2007, Travis Desell wrote:
>
>> Date: Mon, 19 Nov 2007 16:04:52 -0500
>> From: Travis Desell <deselt at cs.rpi.edu>
>> Reply-To: astro at cs.rpi.edu
>> To: astro at cs.rpi.edu
>> Subject: [astro] boinc update
>>
>> Just an update on how things are going on the boinc server.  I ran 5
>> different genetic searches over
>> the weekend with an increased volume (700,1600,80) which is about 8
>> times more than what's usually used, and
>> getting over 3.5 work units returned a second.  Compared to using the
>> large 1024 processor partition of the bluegene,
>> on the smaller volume, which does approximately 2 work units a second -
>> this is pretty impressive.
>>
>> The next set of searches i'm running are using a volume of 700,800,80,
>> so we'll be able to compare if increasing the
>> volume size has any effect on the convergence rate of the search.  I'm
>> working on getting asynchronous particle swarm
>> and simulated annealing searches up and running on boinc as well.   I
>> should also have a that will create new searches
>> from the command line on the milkyway server so Nate and Heidi should be
>> able to create their own searches easily as
>> well if needed.
>>
>> --
>> ----------------------------------------------------------------------------------------------------------
>> Travis Desell
>> <deselt @ cs.rpi.edu>
>> 1-518-878-7424
>> Worldwide Computing Laboratory ( http://wcl.cs.rpi.edu/ )
>> Rensselaer Polytechnic Institute, 110 8th Street, Troy NY 12180, USA
>> ----------------------------------------------------------------------------------------------------------
>>
>>
>>
>
>