What was your usage pattern? This doesn't sound that different than static web file serving, which can be extremely fast. The file system relies heavily on th

What was your usage pattern? This doesn't sound that different than static web file serving, which can be extremely fast. The file system relies heavily on the buffer cache to be fast, and I know there some options to tune that, although I'm not an expert. Do you mean that dir entry lookup is say O(10,000) when you have 10,000 files in a single dir? If so then I imagine that is the reason that git has a hierarchy, i.e. look in .git/objects/[0-f][0-f]/.