From owner-freebsd-chat Wed Sep 3 17:32:00 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id RAA18551 for chat-outgoing; Wed, 3 Sep 1997 17:32:00 -0700 (PDT) Received: from fly.HiWAAY.net (root@fly.HiWAAY.net [208.147.154.56]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id RAA18534 for ; Wed, 3 Sep 1997 17:31:54 -0700 (PDT) Received: from nospam.hiwaay.net (tnt2-7.HiWAAY.net [208.147.148.7]) by fly.HiWAAY.net (8.8.6/8.8.6) with ESMTP id TAA01548 for ; Wed, 3 Sep 1997 19:31:48 -0500 (CDT) Received: from nospam.hiwaay.net (localhost [127.0.0.1]) by nospam.hiwaay.net (8.8.7/8.8.4) with ESMTP id TAA07828 for ; Wed, 3 Sep 1997 19:31:46 -0500 (CDT) Message-Id: <199709040031.TAA07828@nospam.hiwaay.net> X-Mailer: exmh version 2.0zeta 7/24/97 To: chat@FreeBSD.ORG From: dkelly@hiwaay.net Subject: Re: (none) In-reply-to: Message from Sean Eric Fagan of "Wed, 03 Sep 1997 16:40:09 PDT." <199709032340.QAA10588@kithrup.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Date: Wed, 03 Sep 1997 19:31:45 -0500 Sender: owner-freebsd-chat@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk Sean Eric Fagan writes: > > An interesting article. FreeBSD does not come out as well as I would have > liked, though -- better than NT, but not as good as BSD/OS or Linux. (They > don't give any numbers for SCO.) > > The numbers they do give for Linux are surprising, in fact -- it would seem > to indicate that Linux is considerably ahead of FreeBSD 2.2.2 in terms of > performance as a Web server, even with several hundred "simultaneous" > connections. The article says: For example, Linux has a long but fairly straightforward configuration file for adding device support to the kernel, but no tools to optimize kernel performance and very little documentation on how to do so by hand. SCO has about a dozen tools and volumes of documentation for literally every single kernel parameter, which makes up for the lack of source with which to hack out yet another Unix variation. FreeBSD has a single basic configuration file with C define statements; the file is used to parse the kernel source into a source tree for compilation using the Unix program called "make." No fancy configuration utilities, no neat toys. You mean config(8) isn't a fancy configuration utility? It sure isn't make. "... basic configuration file with C define statements..." ? Is he talking about /usr/src/sys/i386/conf/{GENERIC,LINT} ? nospam: {704} cp /usr/src/sys/i386/conf/GENERIC /tmp nospam: {705} cd /tmp nospam: {706} mv GENERIC GENERIC.c nospam: {707} cc GENERIC.c GENERIC.c:8: invalid preprocessing directive name GENERIC.c:60: unterminated character constant GENERIC.c:81: unterminated character constant GENERIC.c:104: unterminated character constant GENERIC.c:157: unterminated character constant nospam: {708} GENERIC didn't look like C to me either. :-) Under the "How We Tested" link: We tested operating systems as shipped, and adjusted performance variables and kernel parameters only when the operating system produced an error condition, such as running out of processes or not having enough memory available to handle the load. Meaning items like "maxusers" was not adjusted? Says their test system had 128 MB of RAM, does FreeBSD report an error in this case? Did they do what was needed so FreeBSD could use all 128M? I went looking for performance charts and missed them if they exist. With Linux running an asynchronous filesystem to get an equal test one would have to add the async option on FreeBSD's mount. Was playing with async this weekend and dropped "make world" times from 2:47 to 2:15 on my new PPro 166/512k. "make depend kernel install" only saved 30 seconds with async, down to 4 minutes 15 seconds. -- David Kelly N4HHE, dkelly@hiwaay.net ===================================================================== The human mind ordinarily operates at only ten percent of its capacity -- the rest is overhead for the operating system.