www.cmm.bristol.ac.uk/forum

Hi Guys,
I am thinking of moving my work on to a cluster? I was wondering if you have had any experience of using runmlwin in a cluster environment? I think the greatest benefits occur when the process are written for paraellel processors. has mlwin be written this way or is it mainly a serial beast.

Model estimation in MLwiN is currently written to use a single processor, however I have heard of cases where lots of instances of MLwiN are run in parallel for example to obtain multiple MCMC chains or conduct a simulation study. If you were going to take this approach then I would suggest calling the engine via a package such as R2MLwiN (http://www.bristol.ac.uk/cmm/software/r2mlwin/) or runmlwin (http://www.bristol.ac.uk/cmm/software/runmlwin/) to allow automation when combining the results.