Subject: Re: Why the partitioning should stay the same
To: Ted Lemon <firstname.lastname@example.org>
From: Greg A. Woods <email@example.com>
Date: 01/31/1995 14:17:29
[ On Tue, January 31, 1995 at 11:15:47 (-0800), Ted Lemon wrote: ]
> Subject: Re: Why the partitioning should stay the same
> Database performance is a very different beast from swap performance.
> One of the primary reasons for going to raw disks is so that you can
> guarantee synchronization. Otherwise you have to use Unix-style
> synchronization, which is comparitively expensive, and generally not a
> good fit for a DBMS. Swap, on the other hand, requires no special
> synchronization, so RDBMS performance requirements really don't apply.
That all depends on what you're measuring. I've seen raw DBMS (i.e. w/o
locking of any sort) change in performance by nearly double when going
to a raw disk vs. UNIX filesystem (Pyramid DC-OS/x & Oracle).
Like it or not, the UNIX filesystem adds a measurable amount of overhead.
Greg A. Woods
+1 416 443-1734 VE3TCP robohack!woods
Planix, Inc. <firstname.lastname@example.org>; UniForum Canada <email@example.com>