It really depends on what you are doing. In general the speed at which you can open a file for reading will be better than the speed at which you can establish a network connection. So for very simple operations, the filesystem is definitely faster. Filesystems will probably beat an RDBMS for raw read throughput too since there is less overhead. In fact, if you think about it, the database can never be faster than the filesystem it sits on in terms of raw throughput.
For very complex operations, the filesystem is likely to be very slow. For example:
Read 10 lines out of this 1 billion line file and then search for matching lines in this other file. I pity you if you have to do this. A good database server however has strategies for doing this fast and well so you aren't reinventing the wheel.
Additionally you really need to figure out what you are doing. What data are you storing? How are you going to transform it? If it is 100k image files your solution will look very different than if it is a directory for 100k people. (LDAP maybe? Or an SQL database? Depends on what you are doing, perhaps.) The key here is to pick the tools that match what you are doing and which give you room to add more uses, rather than whatever seems fastest for some rather abstract use case. Databases are wonderful tools, but you can't get a good answer to a question like this.
Finally premature optimization is the root of all evil. Choose useful tools now and figure out the rest later.