From owner-freebsd-chat Mon Jul 12 18:25:36 1999 Delivered-To: freebsd-chat@freebsd.org Received: from dt054n86.san.rr.com (dt054n86.san.rr.com [24.30.152.134]) by hub.freebsd.org (Postfix) with ESMTP id C80EB15230 for ; Mon, 12 Jul 1999 18:25:19 -0700 (PDT) (envelope-from Doug@gorean.org) Received: from localhost (doug@localhost) by dt054n86.san.rr.com (8.8.8/8.8.8) with ESMTP id SAA09400; Mon, 12 Jul 1999 18:23:00 -0700 (PDT) (envelope-from Doug@gorean.org) Date: Mon, 12 Jul 1999 18:22:59 -0700 (PDT) From: Doug X-Sender: doug@dt054n86.san.rr.com To: David Schwartz Cc: mavery@mail.otherwhen.com, kris@airnet.net, freebsd-chat@freebsd.org Subject: RE: 3C905 versus Intel Etherexpress PRO/100?! In-Reply-To: <000001beccc7$b6f21830$021d85d1@youwant.to> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-chat@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org On Mon, 12 Jul 1999, David Schwartz wrote: > > > > A fundamental design element for a server OS (as opposed to a > > > desktop OS) is to always assume that *every* cpu cycle is valuable. > > Funny, this claim is the opposite of what I've usually heard. Generally, > for desktop use, performance is considered more important that stability or > reliability, which is why people often overclock processors in their > desktops. Well I was thinking more along the lines of, "Sure, I'll put X and massive graphical screensavers on my desktop because I'm the only one who has to wait if that slows it down." As opposed to, "Don't run anything on the server that doesn't absolutely have to be there because if that slows down it affects my customer's perception of my service." Overclocking is an interesting angle to approach this from though. I overclock(ed) my desktop and server processors for my little home system, but I'd never overclock a machine I didn't have daily physical contact with because the possible damage from heat (resulting in big down time) outweighs the possible benefits. > Whereas, for server use, performance (I'm talking 5% or 10%, not factors of > 2) takes a back seat to almost everything else. You can always buy a faster > CPU, or another CPU, for your server. You can always add more RAM. You are > far more concerned with things like clean design, extra safety checks to > prevent crashes, and so on. *Nod* The argument I seemed to be hearing from the poster I responded to was, "Why add that optimization to the code if it only buys us 2%?" He has since corrected my impression. The types of decisions you're talking about are far more likely to come about in terms of, "If we add this sanity check to the code we take a 5% performance hit, but it increases reliability from 98% to 99%. In both the desktop and the server case I'd say go for it, since I don't want avoidable downtime on either machine. This is where sun has freebsd beat cold. They may only be operating at 80% efficiency, but at 99.99% reliability. (Yes I know that there are exceptions to both cases, I'm talking in the general case.) Now that I've given everybody plenty to argue about, I'm going to go buy some pants. Have a nice night, Doug -- On account of being a democracy and run by the people, we are the only nation in the world that has to keep a government four years, no matter what it does. -- Will Rogers To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-chat" in the body of the message