I search an implementation of a network (or distributed) file system like NFS in Java. The goal is to extend it and do some research stuff with it. On the web I found some implementation e.g. DJ NFS, but the open question is how mature and fast they are.
Can anyone purpose a good starting point, has anyone experience with such things?
P.S. I know Hadoop DFS and I used it for some projects, but Hadoop is not a good fit for the things I want to do here.
--EDIT-- Hadoop is really focused on highly scalable, high throughput computing without the possibilities to overwrite parts of a file and so an. The goal is you could use the filesystem e.g. for user home directories.
--EDIT-- More Details: The idea is to modify such a implementation so that the files are not stored directly on a local filesystem, but to apply data de-duplication.