From owner-freebsd-questions Mon Jul 10 11: 9:23 2000 Delivered-To: freebsd-questions@freebsd.org Received: from relay1.inwind.it (relay1.inwind.it [212.141.53.67]) by hub.freebsd.org (Postfix) with ESMTP id 5E27737B6B5 for ; Mon, 10 Jul 2000 11:09:09 -0700 (PDT) (envelope-from bartequi@inwind.it) Received: from bartequi.ottodomain.org (212.141.78.108) by relay1.inwind.it; 10 Jul 2000 20:09:07 +0200 From: Salvo Bartolotta Date: Mon, 10 Jul 2000 19:09:56 GMT Message-ID: <20000710.19095600@bartequi.ottodomain.org> Subject: Re: Softupdates question To: Doug Barton Cc: freebsd-questions@FreeBSD.ORG In-Reply-To: <396A0280.98065D7E@gorean.org> References: <20000709005612.A89313@localhost.localdomain> <20000709.23515500@bartequi.ottodomain.org> <3969172D.D3A30104@gorean.org> <20000710.13040500@bartequi.ottodomain.org> <396A0280.98065D7E@gorean.org> X-Mailer: SuperCalifragilis X-Priority: 3 (Normal) MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG >>>>>>>>>>>>>>>>>> Original Message <<<<<<<<<<<<<<<<<< On 7/10/00, 6:06:08 PM, Doug Barton wrote regarding=20 Re: Softupdates question: > Here is about as clear a post as you're going to get: >=20 http://www.FreeBSD.org/cgi/getmsg.cgi?fetch=3D106939+109577+/usr/local/w= w w/db/text/1998/freebsd-current/19980927.freebsd-current > Try looking at the situation logically. The noatimes option causes > access time not to be written out when a file is accessed. This is=20 part > of the metadata that softupdates optimizes the writing of. Therefore=20 at > best, eliminating that one small piece of metadata is not likely to=20 have > any beneficial effect. > On the other side of the coin, using softupdates & noatime=20 together > used to cause kernel panics. It doesn't cause them anymore, but the=20 code > to work around that bug might be affecting your performance=20 negatively. > I'm not sure on that, but my whole point is that it is so unlikely=20 that > the noatime option is gaining you anything that it's not worth the=20 risk. > Doug > -- Thanks a lot, now I can see your point and your caution :-) In more recent posts, a number of people have made use of softupdates=20 and noatime, and, implicitly or explicitly, they have suggested using=20 the combination of both. This made me (implicitly/automatically) think that the code had been=20 tuned accordingly: that is, it had been designed to work well even=20 with noatime. That's why I have been using both softupdates and=20 noatime for months: I was not aware of the potential risks. =20 What is the state of the art of the matter, if I may ask this forum ? As Paul Herman (and others) have written so far, there seems to be a=20 little improvement in performance of the order of 0.001 up to 0.01 or=20 so. (Ie, of the kind "gutta cavat lapidem"). Are there serious (or any) risks at all nowadays ? =20 Many, many thanks again for your time. Best regards, Salvo To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message