[ On Tue, January 31, 1995 at 17:30:03 (+1030), Brett Lymn wrote: ]
> Subject: Re: Why the partitioning should stay the same
>
> I am not saying that "God hath spake" and made it right but what
> Dr. McKusick said does gel with my experiences on Suns - i.e. that the
> difference was so small I couldn't tell the diff.

I do know you can easily measure the difference under the right
circumstances.  The principle is not much different than that of using
raw disks for an RDBMS like Oracle, and anyone with a sufficiently large
Oracle application can tell you that raw disk access is essential.

I would guess that the kernel overhead of going through the filesystem
and buffer cache is well hidden on systems with poor spindle
organisation, but if you've got decent, fast swap spindles, and enough
other uses of the buffer cache to max it out, you'll notice a remarkable
difference between having filesystems on those swap spindles, and not
having them.

-- 
						Greg A. Woods

+1 416 443-1734			VE3TCP		robohack!woods
Planix, Inc. <woods@planix.com>; UniForum Canada <woods@uniforum.ca>