From: Stephen C. U. <up...@mi...> - 2002-03-20 20:58:57
|
Rick, Well, thank you for the pointer to sun grid engine! Looks like it might be useful as the distributing mechanism, but you're right, not for specifically doing parameter sweeps. We've been using Condor as our underlying distribution mechansim, and the generic experimentation software I've been working on just generates lots of Condor submit jobs, each representing one case, as you've called it. (we call them excursions) steve Rick Riolo wrote: > steve, > > thanks for the pointer to nimrod. > > another tool we have looked at a little is the > sun grid engine (sge). i don't think its specifically > aimed at running sweeps and the like, but its easy > to get going and might be useful. > > - r > > -- > Rick Riolo rl...@um... > Center for Study of Complex Systems (CSCS) > 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, 20 Mar 2002, Stephen C. Upton wrote: > > > Date: Wed, 20 Mar 2002 13:36:08 -0500 > > From: Stephen C. Upton <up...@mi...> > > To: Nick Collier <nic...@ve...> > > Cc: "rep...@li..." > > <rep...@li...> > > Subject: Re: [Repast-interest] batch parameters > > > > Hi, > > > > FWIW, for those in interested in other tools for parameterized experiments, > > you might want to check out this website/tool: > > http://www.csse.monash.edu.au/~davida/nimrod.html/ > > > > >From the webpage: > > Parametric computational experiments are becoming increasingly important in > > science and engineering as > > a means of exploring the behavior of complex systems. For example, an > > engineer may explore the > > behaviour of a wing by running a computational model of the airfoil multiple > > times while varying key > > parameters such as angle of attack, air speed, etc. The results of these > > multiple experiments yield a picture > > of how the wing behaves in different parts of parametric space. > > > > Over the past several years, we have developed a specialized parametric > > modeling system called Nimrod. > > Nimrod uses a simple declarative parametric modeling language to express a > > parametric experiment and > > provides machinery that automates the task of formulating, running, > > monitoring, and collating the results > > from the multiple individual experiments. Equally important, Nimrod > > incorporates a distributed scheduling > > component that can manage the scheduling of individual experiments to idle > > computers in a local area > > network. Together, these features mean that even complex parametric > > experiments can be defined and run > > with little programmer effort. In many cases it is possible to establish a > > new experiment in minutes. Nimrod > > has been applied to a range of application areas, including Bioinformatics, > > Operations Research, Network > > Simulation, Electronic CAD, Ecological Modelling and Business Process > > Simulation. > > > > Not sure how much work it would be to use this, though. Looks like there may > > also be some funky dependencies/reliance on EnFuzion. May be a source for > > ideas, too!! > > > > > > steve > > > > Nick Collier wrote: > > > > > Hi, > > > > > > While I was away, there was quite a bit of discussion about parameter > > > files and running repast in batch mode. I've read over it and there's > > > quite a bit there. So, if possible, I'd like to hear from y'all again on > > > what you you'd like to see, and what, if anything, you are working on in > > > this area. If you are working on something, it would good to indicate > > > whether this is something that could be included in repast in the > > > future. (This will hopefully avoid duplicated effort.) Please cc the > > > list so we can all see what's going on. > > > > > > At the moment, you can run repast in batch mode, through a command line > > > switch and feeding it a parameter file. The parameter file is "dumb" in > > > that it defines a parameter space and iterates over every element in > > > that space. There is no provision for algorithmic exploration of > > > "interesting" parameter spaces. When run in batch mode, recording > > > mechanisms like the DataRecorder record all the data to a single file, > > > recording the batch run number and the dynamic parameter values for that > > > run. More specifically, dynamic parameters (as defined in the parameter > > > file), and the run number are written to the "data table" together with > > > user defined data while static parameters (i.e. those that don't change > > > over the course all the runs) are defined in the header. > > > > > > The parameter file format is a custom one with lots of silly brackets. > > > There is also a little known and undocumented XML format that used to > > > work, but I haven't tested it in a while. In general you can define > > > ranges of values and how to increment through those ranges, as well as > > > lists of numbers, booleans and strings. Single static numeric, boolean, > > > or string values are also accepted. > > > > > > This is the default way to run repast in batch mode. I think Stephen > > > Upton is working on a more general way to do batch runs where the "run > > > mechanism" is more separate from repast itself. Anyway, comments, > > > replies, suggestions appreciated. > > > > > > thanks, > > > > > > Nick > > > > > > -- > > > Nick Collier > > > Social Science Research Computing > > > University of Chicago > > > http://repast.sourceforge.net > > > > > > _______________________________________________ > > > Repast-interest mailing list > > > Rep...@li... > > > https://lists.sourceforge.net/lists/listinfo/repast-interest > > > > > > > > _______________________________________________ > > Repast-interest mailing list > > Rep...@li... > > https://lists.sourceforge.net/lists/listinfo/repast-interest > > |