Network analysis

ingLuca · Post by **ingLuca** » Wed Dec 19, 2007 4:26 am

Having obtained no answer in the User section I try to post here my question

It's possible to run OpenSEES over a small network? In order to speed up the analysis, expecially for nonlinear time history wich takes a lot of time..

Thanks

Post by **fmk** » Wed Dec 19, 2007 10:53 am

opensees does run on parallel machines. i am working on a release for single multi-processor
windows machines. this in theory will also work on a cluster providing you have mpich set up to
work on the cluster.

a parallel version based solely on TCP is also possible. However the parallel part would be limited to element state determination, the matrix equations would be solved sequentially as all the
parallel solvers that we have are written assuming mpi communication.

ingLuca · Post by **ingLuca** » Wed Dec 19, 2007 1:49 pm

Thank you for your reply, but I'm not so inside to software programming to understand all the words you said

Does it mean that I, as an elementary user, can't run the parallel analysis? Or it means that I can run on different machines, but without a big time saving? How could I do? Can you explain in simpler words?

Thanks a lot

Post by **fmk** » Thu Dec 20, 2007 10:54 am

i suggest you use google to find out and learn about mpi and the mpich distribution of it and how to run parallel .exe with it.

koduru · Post by **koduru** » Tue Jan 01, 2008 8:01 pm

Hi fmk,

It would be great if OpenSees utilizes all processors in multi-core PCs.

Are there any future plans to parallelise reliability code too?
Some reliability methods could easily lend themselves to this.

Thanks,