AMBER Archive (2009)

Subject: Re: [AMBER] Use of Amber in distributed calculations

From: M. L. Dodson (activesitedynamics_at_comcast.net)
Date: Sat Jan 17 2009 - 07:48:42 CST

On Sat, Jan 17, 2009 at 10:33:55AM +0300, drugdesign wrote:
> Dear Amber users, I am making a review on molecular dynamics software, which can be used for distributed calculations for the project like FOLDING_at_HOME.
> I wonder whether there is Amber client version which can be used for the
> distributed calculations like boinc or F_at_Home software?
> Don't you by chance know some @home projects which use amber client for
> molecular dynamics?
>
> Best regards,
> Andrew
>

The Amber developers will probably have something to say about this,
but this kind of parallelism won't work with Amber (at least as it is
generally used). Folding_at_home works because each calculation
"quantum" is almost completely independent. Only the final results
need to be communicated back to a central location. The parallelism
in Amber is MUCH, MUCH, MUCH tighter. Even a dedicated Gbit ethernet
interconnect is not fast enough to get good scaling given the speed of
modern "PC" computers. There has been an enormous amount of
discussion on this list about this issue. Use the archive at the
Amber web site to search for the threads.

Bud Dodson

PS, This situation is unfortunate for those of us that would like to
use networks of cheap PC workstations to speed up our calculations.
Others may have things to say about this, or know of special
situations where distributed calculations DO speed things up using
Amber. Obviously, I'm not aware of any. 

-- 
M. L. Dodson
Business Email: activesitedynamics-at-comcast-dot-net
Personal Email:	mldodson-at-comcast-dot-net
Phone:	eight_three_two-56_three-386_one

_______________________________________________
AMBER mailing list
AMBER_at_ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber