From owner-freebsd-arch Thu May 17 7:25:40 2001 Delivered-To: freebsd-arch@freebsd.org Received: from point.osg.gov.bc.ca (point.osg.gov.bc.ca [142.32.102.44]) by hub.freebsd.org (Postfix) with ESMTP id E1B7137B422 for ; Thu, 17 May 2001 07:25:36 -0700 (PDT) (envelope-from Cy.Schubert@uumail.gov.bc.ca) Received: (from daemon@localhost) by point.osg.gov.bc.ca (8.8.7/8.8.8) id HAA00166; Thu, 17 May 2001 07:24:37 -0700 Received: from passer.osg.gov.bc.ca(142.32.110.29) via SMTP by point.osg.gov.bc.ca, id smtpda00163; Thu May 17 07:24:28 2001 Received: (from uucp@localhost) by passer.osg.gov.bc.ca (8.11.2/8.9.1) id f4HEONI68008; Thu, 17 May 2001 07:24:23 -0700 (PDT) Received: from cwsys9.cwsent.com(10.2.2.1), claiming to be "cwsys.cwsent.com" via SMTP by passer9.cwsent.com, id smtpdN67990; Thu May 17 07:24:05 2001 Received: (from uucp@localhost) by cwsys.cwsent.com (8.11.3/8.9.1) id f4HEO4g05590; Thu, 17 May 2001 07:24:04 -0700 (PDT) Message-Id: <200105171424.f4HEO4g05590@cwsys.cwsent.com> Received: from localhost.cwsent.com(127.0.0.1), claiming to be "cwsys" via SMTP by localhost.cwsent.com, id smtpdHu5586; Thu May 17 07:23:15 2001 X-Mailer: exmh version 2.3.1 01/18/2001 with nmh-1.0.4 Reply-To: Cy Schubert - ITSD Open Systems Group From: Cy Schubert - ITSD Open Systems Group X-Sender: schubert To: Tor.Egge@fast.no Cc: dillon@earth.backplane.com, arch@FreeBSD.ORG Subject: Re: on load control / process swapping In-reply-to: Your message of "Thu, 17 May 2001 01:31:06 +0200." <200105162331.BAA04708@midten.fast.no> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Date: Thu, 17 May 2001 07:23:15 -0700 Sender: owner-freebsd-arch@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG In message <200105162331.BAA04708@midten.fast.no>, Tor.Egge@fast.no writes: > ----Next_Part(Thu_May_17_01:30:16_2001)-- > Content-Type: Text/Plain; charset=us-ascii > Content-Transfer-Encoding: 7bit > > > I'd have to see your test code. Doing a direct-read into a user buffer > > has no cache impact at all (DMA does not go through the cpu cache). > > If you are doing seek/read()s but not actually looking at the data that > > is returned, your test results are going to be seriously skewed. > > The test code does not look at the data. I sent a copy of it to you > at January 7th 2000 (along with a previous version of the O_DIRECT > patch). > > I agree that the 95% reduction in CPU usage is seriously skewed. The > performance improvement for most real applications will be very small > or even negative. For some specialized applications it is a > significant performance improvement, giving nearly the same > performance as when bypassing the kernel file system and using the raw > device directly. Specialised applications such as a DBMS using its own cache, e.g. Oracle with a large SGA, would benefit from this. VxFS and Solaris UFS (since Solaris 2.6) have an option to turn on directio for all objects on a filesystem. Assuming you have a disk with only tablespaces on it (quite common in an Oracle shop), applying directio to a whole filesystem is not as damaging to other I/O as one would think because there would be no other I/O to the disk, only Oracle I/O. Regards, Phone: (250)387-8437 Cy Schubert Fax: (250)387-5766 Team Leader, Sun/Alpha Team Internet: Cy.Schubert@osg.gov.bc.ca Open Systems Group, ITSD, ISTA Province of BC To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-arch" in the body of the message