Re: installing emc on a server questions

[Fred Proctor <frederick.proctor-at-nist.gov>]

I run the EMC from an NFS mount all the time. File access is
theoretically slower, but I don't notice it. Once the executables load
into memory, the NFS disk images of the executables aren't accessed.

I do notice that compiles run slower, since all the files are pulled off
the NFS mount, compiled locally, and written back out to the NFS drive.
You'll rarely if ever do compiles, so this isn't an issue.

When running multiple EMCs from the same executables (e.g.,
bridgeportio, stgmod.o), regardless of NFS mounts or local filesystems,
each of the 6 machines has to have its own copy of each of these,
suitable renamed:

generic.ini
generic.nml
generic.var
generic.tbl
generic.run

I recommend this:

1. In the NFS-mounted directory .../emc, create uniquely-named copies of
the generic.ini files, e.g., machine1.ini, machine2.ini, etc. Do the
same for the .nml, .var, and .tbl files. In each of the .ini files, you
specify the name of the .nml, .var, and .tbl files, e.g, in machine1.ini
you change emc.nml to machine1.nml, emc.tbl to machine1.tbl, and emc.var
to machine1.var. Similarly for the other machines. You'll now have a
bunch of files on the NFS mount.

2. For generic.run, create machine1.run, machine2.run, etc. as
executable scripts that just call generic.run with the appropriate .ini
file, like this:

---file machine1.run---

#!/bin/bash

exec ./generic.run -ini machine1.ini
exit 0

---end of file---

Do the same with machine2.run, with -ini machine2.ini, etc.

--Fred