Path: utzoo!utgpu!cs.utexas.edu!usc!samsung!rex!uflorida!mthvax!screamer!tscs!daver!indetech!jack!mvp From: mvp@jack.sns.com (Mike Van Pelt) Newsgroups: alt.hackers Subject: Re: Speaking of load averages... Message-ID: <1991Apr5.032521.3849@jack.sns.com> Date: 5 Apr 91 03:25:21 GMT References: Distribution: alt Organization: Systems'n'Software Lines: 24 Approved: by me. In article bagchi@eecs.umich.edu (Ranjan Bagchi) writes: >In article afoiani@nmsu.edu (Anthony "Tkil" Foiani) writes: >>Here's one from last semester sometime: >> >>8:21am up 4 days, 23:12, 5 users, load average: 279.77, 278.75, 270.68 > My God!...what in hell were these 5 users running? I'd like >to see the ps table... I'm not sure of the nitty-gritty details of how Unix calculates "load averages", but there's bound to be a certain amount of "Finagle factors" in there. I had a program on a Univac 1100 (a.k.a. Sperry, a.k.a. Unisys) that used up about 30-60 CPU "minutes" per second. In that case, certain EXEC requests were "billed" a fixed amount of CPU time, even if they failed instantly because the request was (deliberately, in this case) incorrectly formatted. "Worst Case" (which, of course, is what I picked for the purpose) was an @BRKPT request via ER CSF$. 35 activities of 'em. Moral of the story -- those numbers are more or less fictitious. -- Caution: Thermostellar device. Mike Van Pelt Handle with care. Headland Technology Do not expose to first-semester (was: Video Seven) Philosophy students. ...ames!vsi1!v7fs1!mvp