filesystems and file access effeciency

say i have a rather deep directory structure which stores stuff (templates, flat files, images, pdfs, etc). The directory tree could traverse 20+ directories deep in some places, but typically content that is accessed regularly is stored within the first 5 levels.

Now, at first this may sound like a big tangled web, but it is all stored in a very logical manner. It's also all served up through a file output script which can determine the directory path to any file based on a simple algorithm, so theres no uberlong links in web pages.

But anyways, what i'm wondering is, can this cause a big performance hit? I'm not an expert on filesystems, but say someone links a file 25 directories deep on a high-traffic site. Now, every time this file is accessed, the server has to find directory a, then search directory a for dir b, then search dir b for dir c, then search dir c for dir d, etc. Am I correct in assuming a filesystem works this way, like a linked list where you must follow each link in the list to get to the end? or does the OS determine through less taxing means the physical location of a file?

On the other hand, say everything, thousands or tens-of-thousands of files, are stored in a few directories. Now, rather than searching 20-30 directories that are relatively empty, one or two huge lists of files have to be searched for every single file request.

I'm thinking in most cases any effeciency hit will be miniscule, as filesystems have been tweaked for a long time now. I'd guess some sort of filesystem cache is in place in most OS's, such as if the file /a/b/c/d/e/.../x/y/z/myfile.pdf is requested numorous times, the OS 'remembers' where that file is without looking it up. that is only a guess though.

I think i'll do some benchmark tests, but any thoughts would be appreciated.