Hi Glen, et al.,

Another idea is to use JADE, or some other similar MAS (multi-agent  
system). There is no simulation infrastructure since it was built as  
middleware for constructing MAS's. However, the basics are there for  
having agents on different machines, migration of agents,  
communication between agents, etc.. It would be interesting if the  
ABM community got together with the MAS community to share ideas, if  
they haven't done so already. I haven't seen much. These agents are  
typically more heavyweight than agents used in ABM/IBM's, e.g.,  
having an expert system as a decision making component, but if you're  
headed that route, it might be an avenue to look into.

I'd be curious if someone has attempted this. It's been on my back  
burner for a while, just not enough critical mass to work on it  
myself.. ;-(

BTW, for #2, we've been using Condor () and some software that  
interfaces with design of experiment techniques to generate the  
experiments, using XML as an input format for the ABM. The software  
takes the experimental design via csv and, makes changes to the XML,  
and then creates the jobs for Condor. As long as your input is in XML  
and you have a command line executable, it works for pretty much any  
ABM. We have it running for a MASON developed sim, Netlogo sims, and  
some other military specific sims.

HTH
steve

On Dec 6, 2007, at 10:02 PM, Glen E. P. Ropella wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Maciej M. Latek on 12/06/2007 05:48 PM:
>> That is very interesting issue you raise. I understand you are  
>> interested in
>> fine-grained distributed memory applications: I have experimented  
>> recently
>> with Proactive.  While learning curve might be steep (which is not  
>> helped by
>> [...]
>> What scale of simulation are we talking about that a decent 8 core  
>> desktop
>> with plenty of RAM is not enough?
>
> Thanks for the lead to Proactive.  I'll take a look at it.
>
> Most of my work these days is in biological modeling, which means  
> ~100s
> of thousands of relatively homogenous agents per simulation.  Coarse
> grained simulation (the kind manageable by "shell scripts and rsh",
> though I prefer MPI) has worked for me so far.  But, I'm about to  
> embark
> on a project that might require a much more fine granularity, or more
> accurately multi-grain hierarchical agents.  I'm just looking for an
> evolutionary path to bridge from the current grain to the finer grain.
>
> - --
> glen e. p. ropella, 971-219-3846, http://tempusdictum.com
> Power never takes a back step - only in the face of more power. --  
> Malcolm X
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.6 (GNU/Linux)
> Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
>
> iD8DBQFHWLfZZeB+vOTnLkoRAuIxAKCg8oGqUfo881AJTeinTVTh3Zb27wCeKgAe
> KnGoFIT4rxU2C00bXKS70UQ=
> =8x99
> -----END PGP SIGNATURE-----