[vlovich123]

I think we’re mostly on the same page. I’m more approaching it from intrinsic complexity vs not. Could a database do a better job than the FS Managing extents manually? Maybe. But I’m not as sold that the win is significant enough vs other techniques that could be taken. So yes the papers should definitely consider the filesystem and they don’t. I’m just not convinced that the fragmentation issue is a filesystem vs db but more of a block allocation thing that would always be there and probably work very similarly with little in the way of optimization. Certainly I’ve seen people too ready to throw away the filesystem without actually confirming the cost benefit.