Xref: utzoo news.software.b:8006 news.admin:14719 Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!elroy.jpl.nasa.gov!swrinde!zaphod.mps.ohio-state.edu!wuarchive!uunet!bfmny0!tneff From: tneff@bfmny0.BFM.COM (Tom Neff) Newsgroups: news.software.b,news.admin Subject: Re: Really funny jokes being missed Message-ID: <57314277@bfmny0.BFM.COM> Date: 28 May 91 01:05:04 GMT References: <1991May20.204313.25340@decuac.dec.com> <17898@celit.fps.com> Reply-To: tneff@bfmny0.BFM.COM (Tom Neff) Followup-To: news.software.b Lines: 26 In article <17898@celit.fps.com> billd@fps.com (Bill Davidson) writes: >In article <1991May20.204313.25340@decuac.dec.com> mjr@hussar.dco.dec.com (Marcus J. Ranum) writes: >>I personally don't care if my server spends 5% more time a day processing >>news, as long as I don't have to spend any more time dealing with it. > >Some of us do. Some of us are lucky to get a computer that can handle >the load even under ideal software conditions. Given that load tends to expand to consume available resources, it is not surprising that a lot of people are already maxed out -- but it means that it's pointless to worry about whether everyone's going to be able to afford a 5% cycle hike for some useful new feature. There will always be people who can't, or have convinced themselves they can't, whether rationally or irrationally. In my experience, 3/4 of the 'maxed out' sites in the world are doing at least one thing so STUPIDLY that they could free up giant chunks of capacity with a stroke, if only they knew what they were doing. I am sure Bill Davidson is not one of these; yet even in the case where a smart admin is managing a system at capacity, there is a tradeoff you can outline. Is building a more error free network worth sacrificing X percent of your throughput every day? If not, exactly WHY not? If some rampaging news explosion that *could* have been prevented by additional "insurance" software -- at a 1% cost per day -- succeeds in knocking your cluster out for a weekday, was it worth it? You can buy your way out of a capacity bind, but not (in general) a reliability bind.