[OpenAFS] Re: why afs backup is so poorly supported

Russ Allbery rra@stanford.edu
Tue, 10 Oct 2006 14:24:35 -0700

Jeffrey Hutzelman <jhutz@cmu.edu> writes:

> On the other hand, usenet servers have soundly beaten us to taking the
> next logical step, which is to punt entirely on using a general-purpose
> filesystem and instead use a purpose-built data store.

Usenet servers have a much easier problem, though, since Usenet articles
are write-once, read-many, and only live for a limited time.  So the
custom data store for a Usenet server is basically a cyclical buffer.  You
start writing articles at the front and store the offset of each article
in an index, and when you reach the end, you start writing at the front
again and invalidate the articles as you overwrite them.

Works great for something like Usenet that's inherently transitory, but
a real file system will require a much broader range of semantics.

Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>