From owner-freebsd-questions Mon Apr 26 8:10: 5 1999 Delivered-To: freebsd-questions@freebsd.org Received: from snowdrop.csv.warwick.ac.uk (snowdrop.csv.warwick.ac.uk [137.205.192.31]) by hub.freebsd.org (Postfix) with ESMTP id 3C03C1514A for ; Mon, 26 Apr 1999 08:09:52 -0700 (PDT) (envelope-from csubl@csv.warwick.ac.uk) Received: from primrose.csv.warwick.ac.uk (primrose [137.205.192.26]) by snowdrop.csv.warwick.ac.uk (8.9.1b+Sun/8.9.1) with ESMTP id QAA25714; Mon, 26 Apr 1999 16:09:49 +0100 (BST) Received: (from csubl@localhost) by primrose.csv.warwick.ac.uk (8.9.1/8.9.1) id QAA02883; Mon, 26 Apr 1999 16:08:28 +0100 (BST) From: Mr M P Searle Message-Id: <199904261508.QAA02883@primrose.csv.warwick.ac.uk> Subject: Re: writing slower with SoftUpdates In-Reply-To: <7ftnto$top$1@mips.rhein-neckar.de> from Christian Weisgerber at "Apr 25, 99 02:34:00 am" To: naddy@mips.rhein-neckar.de (Christian Weisgerber) Date: Mon, 26 Apr 1999 16:08:28 +0100 (BST) Cc: freebsd-questions@freebsd.org X-Mailer: ELM [version 2.4ME+ PL31 (25)] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG > Dan Nelson wrote: > > > If these are in fact identical drives, I would start looking at the > > mode pages on the first drive to see if you have write caching turned > > on. Softupdates usually prefers to cache data itself, and you can get > > degraded performance if the drive is doing caching too. > > In what way do soft updates perform additional caching? Writes are not made immediately, but are delayed in the same way that a HD cache does. > And why should this interact badly with drive caching? I don't know, but the drive cache doesn't seem to make any difference to the speed when using soft updates. (the HD's cache is so much smaller that the amount of memory buffers soft updates can use it won't make any difference.) > Disabling the > drive's write caching seems like an excellent way to cripple the drive > to me. > Soft updates is also not just a dumb cache like the HD cache, but knows about the file system structure. This makes it much safer than running either the old async file system or a HD cache as the file system is unlikely to be seriously corrupted in a crash. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message