From owner-freebsd-performance@FreeBSD.ORG Sun Jul 16 14:49:31 2006 Return-Path: X-Original-To: freebsd-performance@freebsd.org Delivered-To: freebsd-performance@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 7F5C616A4DE for ; Sun, 16 Jul 2006 14:49:31 +0000 (UTC) (envelope-from freebsd@hub.org) Received: from hub.org (hub.org [200.46.204.220]) by mx1.FreeBSD.org (Postfix) with ESMTP id 14D6643D45 for ; Sun, 16 Jul 2006 14:49:31 +0000 (GMT) (envelope-from freebsd@hub.org) Received: from localhost (mx1.hub.org [200.46.208.251]) by hub.org (Postfix) with ESMTP id 8410C290C6B for ; Sun, 16 Jul 2006 11:49:24 -0300 (ADT) Received: from hub.org ([200.46.204.220]) by localhost (mx1.hub.org [200.46.208.251]) (amavisd-new, port 10024) with ESMTP id 12330-04 for ; Sun, 16 Jul 2006 11:49:30 -0300 (ADT) Received: from ganymede.hub.org (blk-224-179-167.eastlink.ca [24.224.179.167]) by hub.org (Postfix) with ESMTP id 0C794290C38 for ; Sun, 16 Jul 2006 11:49:24 -0300 (ADT) Received: by ganymede.hub.org (Postfix, from userid 1027) id 0ADD14882D; Sun, 16 Jul 2006 11:49:36 -0300 (ADT) Received: from localhost (localhost [127.0.0.1]) by ganymede.hub.org (Postfix) with ESMTP id 035AE3BC76 for ; Sun, 16 Jul 2006 11:49:36 -0300 (ADT) Date: Sun, 16 Jul 2006 11:49:35 -0300 (ADT) From: User Freebsd To: freebsd-performance@freebsd.org Message-ID: <20060716114546.B1799@ganymede.hub.org> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed Subject: Is 6.x slower then 4.x ... ? X-BeenThere: freebsd-performance@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Performance/tuning List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sun, 16 Jul 2006 14:49:31 -0000 I've read/seen reports on -questions about this ... especially in an SMP environment ... Is there any truth to this? One person that posted on -questions pointed out that when he tried to point out the difference, he was told one of: > a. It is either your hardware sucks > b. your benchmark application sucks Now, I don't hold with the 'your hardware sucks' response, as long as one is testing on the same hardware, the hardware itself should affect the results between releases ... But, is there any "officially recognized set of tests" that one can use that ppl here accept to negate the chances of b? Finally, has anyone here done a set of 'accepted tests' and built up a report that could be linked to from the main page to refute (or vindicate?) the claims that the newer releases are getting slower? ---- Marc G. Fournier Hub.Org Networking Services (http://www.hub.org) Email . scrappy@hub.org MSN . scrappy@hub.org Yahoo . yscrappy Skype: hub.org ICQ . 7615664 From owner-freebsd-performance@FreeBSD.ORG Mon Jul 17 03:03:01 2006 Return-Path: X-Original-To: freebsd-performance@freebsd.org Delivered-To: freebsd-performance@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 1581816A4DD for ; Mon, 17 Jul 2006 03:03:01 +0000 (UTC) (envelope-from brooks@lor.one-eyed-alien.net) Received: from sccmmhc92.asp.att.net (sccmmhc92.asp.att.net [204.127.203.212]) by mx1.FreeBSD.org (Postfix) with ESMTP id B010B43D58 for ; Mon, 17 Jul 2006 03:02:58 +0000 (GMT) (envelope-from brooks@lor.one-eyed-alien.net) Received: from lor.one-eyed-alien.net ([12.207.12.9]) by sccmmhc92.asp.att.net (sccmmhc92) with ESMTP id <20060717030256m92002t4m4e>; Mon, 17 Jul 2006 03:02:57 +0000 Received: from lor.one-eyed-alien.net (localhost [127.0.0.1]) by lor.one-eyed-alien.net (8.13.6/8.13.6) with ESMTP id k6H32ovR003576; Sun, 16 Jul 2006 22:02:54 -0500 (CDT) (envelope-from brooks@lor.one-eyed-alien.net) Received: (from brooks@localhost) by lor.one-eyed-alien.net (8.13.6/8.13.6/Submit) id k6H32oTo003575; Sun, 16 Jul 2006 22:02:50 -0500 (CDT) (envelope-from brooks) Date: Sun, 16 Jul 2006 22:02:50 -0500 From: Brooks Davis To: User Freebsd Message-ID: <20060717030249.GB3344@lor.one-eyed-alien.net> References: <20060716114546.B1799@ganymede.hub.org> Mime-Version: 1.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="NMuMz9nt05w80d4+" Content-Disposition: inline In-Reply-To: <20060716114546.B1799@ganymede.hub.org> User-Agent: Mutt/1.5.11 Cc: freebsd-performance@freebsd.org Subject: Re: Is 6.x slower then 4.x ... ? X-BeenThere: freebsd-performance@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Performance/tuning List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 17 Jul 2006 03:03:01 -0000 --NMuMz9nt05w80d4+ Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable On Sun, Jul 16, 2006 at 11:49:35AM -0300, User Freebsd wrote: >=20 > I've read/seen reports on -questions about this ... especially in an SMP= =20 > environment ... >=20 > Is there any truth to this? >=20 > One person that posted on -questions pointed out that when he tried to=20 > point out the difference, he was told one of: >=20 > >a. It is either your hardware sucks > >b. your benchmark application sucks >=20 > Now, I don't hold with the 'your hardware sucks' response, as long as one= =20 > is testing on the same hardware, the hardware itself should affect the=20 > results between releases ... >=20 > But, is there any "officially recognized set of tests" that one can use= =20 > that ppl here accept to negate the chances of b? >=20 > Finally, has anyone here done a set of 'accepted tests' and built up a=20 > report that could be linked to from the main page to refute (or=20 > vindicate?) the claims that the newer releases are getting slower? FreeBSD 6 is slower than 4 for some things and faster for others. That should be expected since fine grained locking involves increased numbers of expensive atomic operations (which are particularly bad on Intel P4 and Xeon systems). The gain is that we've got significantly more parallelism in many areas (for example, see kris's I/O benchmarking presented at BSDCan). Looking at it as a thought experiment, you should expect microbenchmarks to perform worse, sometimes much worse. If your application looks like those microbenchmarks that's going to be a problem, if not it may or may not be. In short the black and white question you are asking makes little sense. :) -- Brooks --NMuMz9nt05w80d4+ Content-Type: application/pgp-signature Content-Disposition: inline -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.2 (FreeBSD) iD8DBQFEuv3ZXY6L6fI4GtQRAnjWAKCSf54AfVBGuWLBCV/zTqwDjH38NACfX16V rY3hYPCYgofQXxJ2s0hmGa0= =JYG1 -----END PGP SIGNATURE----- --NMuMz9nt05w80d4+-- From owner-freebsd-performance@FreeBSD.ORG Mon Jul 17 09:09:58 2006 Return-Path: X-Original-To: freebsd-performance@freebsd.org Delivered-To: freebsd-performance@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id A729916A4DA for ; Mon, 17 Jul 2006 09:09:58 +0000 (UTC) (envelope-from ah@shadow.splashground.de) Received: from shadow.splashground.de (shadow.splashground.de [85.214.34.252]) by mx1.FreeBSD.org (Postfix) with SMTP id DE97643D6A for ; Mon, 17 Jul 2006 09:09:57 +0000 (GMT) (envelope-from ah@shadow.splashground.de) Received: (qmail 23391 invoked by uid 1001); 17 Jul 2006 11:09:57 +0200 Date: 17 Jul 2006 11:09:57 +0200 Message-ID: <20060717090957.23390.qmail@shadow.splashground.de> User-Agent: Emai/0.0.6 Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Disposition: inline References: <20060716114546.B1799@ganymede.hub.org> <20060717030249.GB3344@lor.one-eyed-alien.net> In-Reply-To: <20060717030249.GB3344@lor.one-eyed-alien.net> From: Andreas Hauser To: Brooks Davis X-License: BSD X-Addicted: yeah Cc: freebsd-performance@freebsd.org, User Freebsd Subject: Re: Is 6.x slower then 4.x ... ? X-BeenThere: freebsd-performance@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Performance/tuning List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 17 Jul 2006 09:09:58 -0000 brooks wrote @ Sun, 16 Jul 2006 22:02:50 -0500: > On Sun, Jul 16, 2006 at 11:49:35AM -0300, User Freebsd wrote: > FreeBSD 6 is slower than 4 for some things and faster for others. That > should be expected since fine grained locking involves increased numbers > of expensive atomic operations (which are particularly bad on Intel > P4 and Xeon systems). The gain is that we've got significantly more > parallelism in many areas (for example, see kris's I/O benchmarking > presented at BSDCan). Looking at it as a thought experiment, you should > expect microbenchmarks to perform worse, sometimes much worse. If > your application looks like those microbenchmarks that's going to be a > problem, if not it may or may not be. OK. Kris presented exactly one benchmark were 6 is better (30%) and that is with sync mounts. Sorry, but i don't know many people running async mounts. Since none of the benchmarks from people seem to have influence on you, why not provide benchmarks, application ones, that show that 6 is good in anything performance wise. Until then we keep thinking it is worse, since our benchmarking shows it to be worse (of course we are doing it all wrong ...). > In short the black and white question you are asking makes little > sense. :) It usually boils down to a black and white question like "Use or do not use?". -- Andy From owner-freebsd-performance@FreeBSD.ORG Mon Jul 17 09:07:33 2006 Return-Path: X-Original-To: freebsd-performance@freebsd.org Delivered-To: freebsd-performance@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 9AE5C16A4F2 for ; Mon, 17 Jul 2006 09:07:33 +0000 (UTC) (envelope-from ah@shadow.splashground.de) Received: from shadow.splashground.de (shadow.splashground.de [85.214.34.252]) by mx1.FreeBSD.org (Postfix) with SMTP id 813B143D72 for ; Mon, 17 Jul 2006 09:07:12 +0000 (GMT) (envelope-from ah@shadow.splashground.de) Received: (qmail 22835 invoked by uid 1001); 17 Jul 2006 11:07:10 +0200 Date: 17 Jul 2006 11:07:10 +0200 Message-ID: <20060717090710.22834.qmail@shadow.splashground.de> User-Agent: Emai/0.0.6 Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Disposition: inline References: <20060716114546.B1799@ganymede.hub.org> <20060717030249.GB3344@lor.one-eyed-alien.net> In-Reply-To: <20060717030249.GB3344@lor.one-eyed-alien.net> From: Andreas Hauser To: Brooks Davis X-License: BSD X-Addicted: yeah X-Mailman-Approved-At: Mon, 17 Jul 2006 12:00:08 +0000 Cc: freebsd-performance@freebsd.org, User Freebsd Subject: Re: Is 6.x slower then 4.x ... ? X-BeenThere: freebsd-performance@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Performance/tuning List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 17 Jul 2006 09:07:33 -0000 brooks wrote @ Sun, 16 Jul 2006 22:02:50 -0500: > On Sun, Jul 16, 2006 at 11:49:35AM -0300, User Freebsd wrote: > FreeBSD 6 is slower than 4 for some things and faster for others. That > should be expected since fine grained locking involves increased numbers > of expensive atomic operations (which are particularly bad on Intel > P4 and Xeon systems). The gain is that we've got significantly more > parallelism in many areas (for example, see kris's I/O benchmarking > presented at BSDCan). Looking at it as a thought experiment, you should > expect microbenchmarks to perform worse, sometimes much worse. If > your application looks like those microbenchmarks that's going to be a > problem, if not it may or may not be. OK. Kris presented exactly one benchmark were 6 is better (30%) and that is with sync mounts. Sorry, but i don't know many people running async mounts. Since none of the benchmarks from people seem to have influence on you, why not provide benchmarks, application ones, that show that 6 is good in anything performance wise. Until then we keep thinking it is worse, since our benchmarking shows it to be worse (of course we are doing it all wrong ...). > In short the black and white question you are asking makes little > sense. :) It usually boils down to a black and white question like "Use or do not use?". -- Andy