From owner-freebsd-hackers Thu Mar 23 19:47:48 1995 Return-Path: hackers-owner Received: (from majordom@localhost) by freefall.cdrom.com (8.6.10/8.6.6) id TAA10249 for hackers-outgoing; Thu, 23 Mar 1995 19:47:48 -0800 Received: from hq.icb.chel.su (icb-rich-gw.icb.chel.su [193.125.10.34]) by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id TAA10221 for ; Thu, 23 Mar 1995 19:46:56 -0800 Received: from localhost (babkin@localhost) by hq.icb.chel.su (8.6.5/8.6.5) id IAA00423; Fri, 24 Mar 1995 08:35:53 -0500 From: "Serge A. Babkin" Message-Id: <199503241335.IAA00423@hq.icb.chel.su> Subject: Re: Why IDE is bad To: chuckr@Glue.umd.edu (Chuck Robey) Date: Fri, 24 Mar 1995 08:35:53 -0500 (GMT-0500) Cc: freebsd-hackers@FreeBSD.org In-Reply-To: from "Chuck Robey" at Mar 23, 95 03:18:17 pm X-Mailer: ELM [version 2.4 PL23] Content-Type: text Content-Length: 2391 Sender: hackers-owner@FreeBSD.org Precedence: bulk > On Thu, 23 Mar 1995, Joe Greco wrote: > > > > > 4) writing the disk more may wear it out faster. > > > > > > Now there's one for the books! :-) > > > > With an 800,000 hour MTBF, I think that enough years will > > have passed that even if I reduce the lifetime of the > > disk by 75%, I'll be too old and grey to care when it > > dies. (yes I am sure the original comment was to see if > > we were all awake :-) but it's a good point). > > I did once see someone wipe a disk clean, by doing continuous > read/write/read cycles on it. The first disk went in 3 months, then > three months later the second disk (and the programmer, by the way) > went. So I think it can happen, but probably not with the kinda load we > see. Some information for everybody who thinks that MTBF is equal to the lifetime: Take an experiment. We take a number of devices and begin to exploit it. We remember when every unit dies until all of them are dead. Then we plot a graph. This graph will be like this (plotted by *s): % dead ^ 100% | * 100%died | * | * | * / --|----------------------------------------* B _| | * | | | * | \ --|------* | | * | | | * | A | | * | _______________|_______________ | |* |/ \| 0 +------+---------------------------------+-----------------> | time of work | time time of lifetime manufacturer's tests and free repair MTBF is measured at time of work when the curve is near linear by dividing A by B: MTBF=A/B Therefore MTBF 800,000 hours doesn't means that HDU will work 800,000 hours before a failure, it only means that if lifetime is near 10,000 hours then near 1/80 of all sold HDUs will die before reaching the lifetime. I'm sorry if I had translated some terms to english wrong. Serge Babkin ! (babkin@hq.icb.chel.su) ! Headquarter of Joint Stock Bank "Chelindbank" ! Chelyabinsk, Russia