Re: Repeating Simulations: Drone or what?

[Rick Riolo]

As you might expect, I use drone for doing batch experiments,
both with swarm and with other programs.

Perhaps a full sim package should have features for running
replicable experiments, including keeping logs of why
one did the experiments, where the control files are for
various experiments, and so on.  
I can see situations where this tight coupling would
be useful.  For example it woudl be useful to automatically monitor 
experiments and start new ones based on the earlier results...this could
generalize into, for example, a system to use an EA (GA) to
search a parameter space.
 
On the other hand, these features are useful and common
to any situation where one wants to do computational experiments.
Thus it would seem to me that it would be useful to have
these experiment running and management features be
separate from any particular program, sim lib, etc.
That's what we've tried to do with drone.
Its not got all I wish it had, but it has proved to be a very good and 
useful tool even as it stands.

I know of one person who is writing a experiment controller/manager
in Java...I don't think its been released yet, and if I get 
some time I would like to work with him to get as much of
drone functionality (actual and wished for) into his package.

You can follow pointers on 
   http://www.pscs.umich.edu/LAB/Doc/SwarmStuff/UM-swarm-index.html
to see how we generally use swarm (and other packages) with drone.
We happen to use a small (kludgy) parameter handling package
for our swarm apps, but (as far as I know) it should be possible
to use the newer parameter processing class built in swarm
to create swarm apps that are drone-compliant.
(Of course we have sven's message about a possible problem...)

Now to show my acronum ignorance:
What are RPMS and SRPMS in the context of the message below? 

 - r

Rick Riolo                           address@hidden
Program for Study of Complex Systems (PSCS)
4477 Randall Lab                
University of Michigan         Ann Arbor MI 48109-1120
Phone: 734 763 3323                  Fax: 734 763 9267
http://www.pscs.umich.edu/PEOPLE/rlr-home.html

On Wed, 26 Aug 1998, Paul Johnson wrote:

> Date: Wed, 26 Aug 1998 08:32:11 -0500 (CDT)
> From: Paul Johnson <address@hidden>
> To: address@hidden
> Subject: Repeating Simulations: Drone or what?
> 
> 
> I'm just for the past few months very interested in making simulations
> repeat and I was excited to learn there is this thing called DRONE
> discussed in Sven's and Rick's recent notes.
> 
> I know it is possible to rewrite a batchswarm to make simulations repeat
> themselves (simpleExerBug) and a perl script can do it.  Drone is new to me, 
> or
> if I knew I had forgotten...
> 
> Here's what I wonder. In the swarm community, how are most people replicating
> experiments?  Are there other approaches I've never heard of?
> 
> Should replication be part of swarm itself? At Swarmfest, Roger described
> the possibility of writing swarm to allow a "reset", so simulations could be
> restarted without quitting.  It seems a small step from that to a repetition.
> 
> It just seems to me that the ability to repeat an exeriment ought not be 
> an "optional added extra" in the simulation package design. Rather, it
> ought to be integral to it.
> 
> Now some shameless advertising! I prepared RPMs and SRPMs for libffi, 
> BLT (your choice: version 2.4e or 8.0-unofficial!) and swarm 1.3. They
> work on both of  my computers and I'm intersted to know if they work on yours.
> I'm interested to know of anyone tries to rebuild the RPMS on SPARC or Alpha
> versions of Redhat too.  
> 
> http://lark.cc.ukans.edu/~pauljohn/Swarm
> 
> 
> Paul E. Johnson                      address@hidden
> Dept. of Political Science           http://lark.cc.ukans.edu/~pauljohn
> University of Kansas                 Office: (913) 864-9086
> Lawrence, Kansas 66045               FAX: (913) 864-5700
>  
> 
> 
>                   ==================================
>    Swarm-Support is for discussion of the technical details of the day
>    to day usage of Swarm.  For list administration needs (esp.
>    [un]subscribing), please send a message to <address@hidden>
>    with "help" in the body of the message.
>                   ==================================
> 

                  ==================================
   Swarm-Support is for discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.
                  ==================================