Shell Programming and Scripting

That its 75 times slower via NFS than disk is a pretty big hint.

Yes, NFS versus disk matters quite a lot. Its asynchronous nature means that an operation you'd consider atomic on a local disk mean a lot of "who has this file? is this file busy? Has anyone updated it since I saw it last? I'm taking this file, LAST WARNING, FILE TAKEN. Everyone who might be reading this file, I have updated it, flush your buffers! I have now CLOSED THE FILE, free for all takers" etc etc messaging back and forth. Imagine the overhead this creates for programs using huge clutters of tiny files! All that asking and waiting. If any of it can be done in parallel that can help, having 3-4 requests flying at once instead of 1, but a shared filesystem is by definition a lot more back-and-forth than a dedicated one.

The traditional answer to "why is my file tree walker slow" is "Is it NFS?" and the traditional reply to that is something along the lines of "Why -- does that matter?" followed by "I configured it to skip the NFS, much better now".

Corona is spot on. Nothing you can do code-wise on the UNIX box will make NFS faster.
You need to short circuit the find and read NFS operations somehow.

If you absolutely have to speed things up no matter what: reverse roles. Locate the files on your UNIX box and then Samba/CIFS/NFS share it to the Windows box.

I found something interesting. *.aps folders are rarely inside other *.aps folders (although *.aps files are . . . just not folders).
folderList=$( find "$dataDir" -type d -name '*.aps' -printf '%T@|%p\n' -prune | sort -nr | cut -d"|" -f 2 ) is not a perfect solution (I miss a low single digit percentage of matches) but it gives me an idea how fast things can be (about 40x faster). Yes, and this proves that NFS is the bottleneck . . .

I found out that if a *.aps folder contains a file called Autorange.lay, you definitely do not need to descend any further to find more *.aps folders.

Mike

If you require 100% accuracy, and can't find a simplification which works 100% of the time, a full search may be unavoidable.