Re: ReiserFS in Standard Kernel?

Edward S. Marshall (emarshal@logic.net)
Wed, 14 Jul 1999 10:05:10 -0500 (CDT)


On Wed, 14 Jul 1999, Horst von Brand wrote:
> "Edward S. Marshall" <emarshal@logic.net> said:
> > Anyone running a reasonably high-traffic news server these days threw out
> > the one-file-per-article storage method a long time ago. It just doesn't
> > scale to 30-40G per day.
>
> But that means extra hair in the server. If the OS can take that over,
> everybody wins, even those that didn't go the extra mile to implement their
> own filesystem since the OS's sort of sucks for their application.

Agreed, but I can still do better with a userspace solution.

Under the one-file-per-article scheme, I have to make at least three
system calls to write an article (open(), write(), close()). Reading a
file is basically the same thing. (You can use mmap() to get around some
performance issues, but many people avoid it for portability reasons.)

Under a scheme with one large file, and the server managing it, I have
two: lseek() and read/write() (open() and close() are done at
initialization and shutdown).

While I don't argue that reiserfs would be a fantastic thing to have (mail
spools, for example, would benefit greatly from it), I just don't see it
as the next big way to do news.

-- 
Edward S. Marshall <emarshal@logic.net>       [ What goes up, must come down. ]
http://www.logic.net/~emarshal/               [ Ask any system administrator. ]

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/