Question on SSH and accessing local files

[jst04004]

There is a program I need to run on Machine B that I access through SSH. This program requires files that I store locally on Machine A. Once done with the program, I need to then get the output files from Machine B back to Machine A.

Once I SSH in, I tried doing this through /Volumes on Machine B, but Machine A is not there. Do I need to use SFTP, or is there a quicker way or any other suggestions that you might have? They would all be greatly appreciated.

[Irithori]

Hi and welcome

you can redirect stdin and stdout from/to MachineA.
But in general, especially if we are talking about more than one input/output file and/or big files,
then I would suggest copying them first.

Can you be more specific about your usecase?

[jst04004]

There is a folder on machine A (group use machine), that contains all the framework and files for the program I want to run. I edit the files as necessary on A, but then need to get them to B which has multiprocessor capabilities. It would take forever without those. Once I have the files in place on B, I go into the run directory within that folder, run the program, and it outputs a few files to the output folder which I need. I need to save those output files as well as the original files that I change to run the simulation. I can only post process the output files on B, which has Matlab and VisIt.

[Irithori]

For mostly cpu bound workloads, nfs (and optionally autofs) comes to mind.

Export the input and output directories of machine A,
and (auto)mount them on machine B.
Machine B now has file access and can start processing.

However, if there is a lot of IO needed, then a local copy might be faster.
So, I would first look at a typical job via atop, iostat, etc with a local copy of input files on machine B.
(This is not a complete benchmark, but should provide a basic idea of the expected workload)