Re: Porting Greg Lehey's rawio.c from FreeBSD to Linux...

Stephen C. Tweedie (sct@redhat.com)
Thu, 6 May 1999 19:38:55 +0100 (BST)


Hi,

On Tue, 4 May 1999 18:58:56 +0200, Brad Knowles <brad@shub-internet.org> said:

> Since sending out that note, I've had this explained to me. Can you
> explain to me the semantics of an Oracle or Sybase database wanting to
> have direct access to a raw disk device for the database, so that:

> 1) You don't trash your buffer cache

Yes

> 2) You don't waste your time and that of the database when it
> could avoid the filesystem redirection entirely?

Yes, and it also avoids the CPU cost of copying data around, so there's
a double performance win there.

> Or am I mistaken in believing that the block device interface is also
> pumped through the vm/buffer cache system?

The block device interface is layered underneath the buffer cache, but
it can be used independently if you fake the buffer_heads. That's what
the VM does when swapping or when populating the page cache, for
example, and the Linux raw character device patches do the same for
block IO.

--Stephen

-
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/