Windows 10 + Linux working together

Specific Domains Julia at Scale
1

I have two computers, one running Windows 10 and the other running Ubuntu 14.04,
both with the same version of the language: Julia 0.7.0.
I would like to combine the computing power of the two computers,
and control a simple distributed pmap calculation from Windows as the master.

Obviously I need a passwordless ssh from Windows to Linux.
For adding processes on Linux, addprocs needs to set up the correct dir and exename.
I suppose I have done these steps correctly.

My code is contained in a single module in a single file.
But as the two filesystems are completely separate,
I suppose I need MyModule.jl to exist on both computers
and I have to synchronize them manually.
– Is this correct?

Then how should I compile and bring them into scope?
– Is “@everywhere workers() using MyModule” the correct way to do that?

– Are there any other gotchas to be expected?

I would greatly appreciate any advice.

2

Technically, the script file doesn’t have to exist twice. It just has to be accessible from both machines. I use sftp and mountain duck to mount a Linux fs in windows.

3 Likes
3

I understand that some kind of file sharing is preferred.

But what if the two file systems remain separate?
Assuming that the source code of MyModule.jl is already synchronized,
how is it achieved that module compilation occurs only once in each of the separate file systems?

As I asked earlier:
– Is “ @everywhere workers() using MyModule ” the correct way to do that?

4

Looks fine to me.