From owner-freebsd-hackers Mon Jan 27 04:45:48 1997 Return-Path: Received: (from root@localhost) by freefall.freebsd.org (8.8.5/8.8.5) id EAA04828 for hackers-outgoing; Mon, 27 Jan 1997 04:45:48 -0800 (PST) Received: from zwei.siemens.at (zwei.siemens.at [193.81.246.12]) by freefall.freebsd.org (8.8.5/8.8.5) with ESMTP id EAA04820 for ; Mon, 27 Jan 1997 04:45:41 -0800 (PST) Received: from sol1.gud.siemens.co.at ([10.1.143.100]) by zwei.siemens.at (8.7.5/8.7.3) with SMTP id NAA19098 for ; Mon, 27 Jan 1997 13:46:24 +0100 (MET) Received: from ws2301.gud.siemens.co.at by sol1.gud.siemens.co.at with smtp (Smail3.1.28.1 #7 for ) id m0voqQL-000213C; Mon, 27 Jan 97 13:44 MET Received: by ws2301.gud.siemens.co.at (1.37.109.16/1.37) id AA114238913; Mon, 27 Jan 1997 13:41:53 +0100 From: "Hr.Ladavac" Message-Id: <199701271241.AA114238913@ws2301.gud.siemens.co.at> Subject: Re: Performance of ufs vs. ext2. To: dicen@hooked.net (dicen) Date: Mon, 27 Jan 1997 13:41:53 +0100 (MEZ) Cc: freebsd-hackers@freebsd.org In-Reply-To: <32EC6CE5.64E60DE1@hooked.net> from "dicen" at Jan 27, 97 00:52:53 am X-Mailer: ELM [version 2.4 PL24 ME8a] Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-hackers@freebsd.org X-Loop: FreeBSD.org Precedence: bulk E-mail message from dicen contained: > Bruce Evans wrote: > > > > >> Have other people tested ufs vs. ext2? The only docs I could find where > > >> ... > > >The performance that I have measured (sequential -- IOZONE) is that > > >FreeBSD is faster in both read/write. However, our metadata performance > > >is slower (filecreates/deletes.) With -async, our metadata is still > > >slower, but not by orders of magnitude. FreeBSD's cache perf is much > > >faster (by factors of 3-4.) Much of it is due to the default block > > >size (8K vs. 1K.) But the fragment size of an 8K UFS filesystem is > > >the *same* as a 1K ext2fs. > > > > In my tests, ext2fs is fastest for huge sequential i/o's when the block > > sizes are closer (8K vs 4K), but there was only a small difference (less > > than 10%) between the best and worst cases (best: ext2fs under FreeBSD, > > next: ext2fs under Linux, worst: ext2fs under Linux) except for rewrite, > > which was 66% faster under Linux than under FreeBSD. Cache performance > > also catches up (46MB/sec for FreeBSD-current-last-November, 41MB/sec > > for Linux-2.0.20). A 4K fragment size wastes space probably wastes time > > in most cases. > > > > Bruce > > Okay cool some real numbers. When you speak of "rewrite" are you talking > about the creation and deletion of files (Metadata)? There seams to be a > significant speed difference between the creation and deletion of files > on linux ext2 vs. Freebsd ufs. Linux ext2 is way faster. I suppose I > could just run ext2 under FreeBSD right? It sure would make a "make > world" faster. You know if someone were to setup a news server it would > seam to make more sence to use ext2. Doubt it. You could run FFS async (the way the Linuxers run ext2fs) and pray to Deity-of-your-choice not to lose the filesystem (the way the Linuxers who run ex2fs should do but do not, presumably out of ignorance). For a news server, you can run FFS noatime, and have a big win without endangering your data. Joe Greco made a series of posts on the matter. /Marino >