Running AnyBody on a computer cluster

Hello Everyone

I am sorry, as I am not sure whether I use a right forum to ask such questions.

I have been working with AnyBody for a while, and using AnyPyTools I would like to run several thousand simulations.

AnyPyTools can run several instances of AnyBody at once, using multithreading to speed up the task. In my working environment (ETH Zurich), we have access to a computer cluster, and using it would significantly increase my computational possibilities.

As I received no information from our IT other than a very short refusal, I wonder what could be the legal/technical obstacles preventing such solution, given that I have an access to the AnyBody on Research Licence.

Thank you very much for any response.

Jakub Sikorski

Hi Jakub

I will leave it to the AnyBody supporters to answer any issues you may have running multiple instances of AnyBody with your license.

But I do have some comments to the technical issues. It may be a problem if your cluster runs linux. I have tried the same here at Aalborg University. I could install AnyBody on Linux using Wine. The GUI didn’t work, but the console (anybodycon.exe) worked fine. I also tried to install it on the university super-computer. However, I ran into some problems where it could not find our license server. It could probably have been solved, but our IT department was not easy to work with. I never got any further, as I didn’t have the need at the time.

We use AnyPyTools to run simulations on a very fast windows workstation. I have never tested AnyPyTools on linux, and it probably needs to be modified a bit work there.

Also, AnyPyTools only has a pretty simple scheduler designed for a single computer with multiple cores. To work well on a super-computer it would need a new scheduler that could use the MPI interface. I have no experience with that, but since there is already a python packages for that (mpi4py), it should be possible to implement even for someone without much experience with MPI.