> > > >However, there is no redundancy in a RAID0 set, hence it's more > commonly > > > >referred to as striping. Actually, introducing RAID0 increases the > > > >probability of a disk crash (and hence data loss without propper > > > >backup) by the increase in disks. > > > > > > What a theory is this?! > > How can it increate a probility of a disk crash? Is it just because of > using > > two disks? > > Is so, it is a nonsense. > > Why is it nonsense? > Suppose the probability of a disk crashing is > "Once in ten years" > > If you do repeated testing using a bank of ten disks, you would expect (on > average) a disk to crash every year. OR, if you're really unlucky, nothing > happens for ten years and then they all explode at exactly the same time at > the end of the tenth year (OBVIOUSLY this is unlikely). But one of the two > must happen - EITHER the probability of "ANY" of the disks crashing is > multiplied by ten because you have ten disks > - i.e. IF you know the probability of a single disk crashing is "once > in x years" then the probability of any one of "n" disks crashing is "once > in (x/n) years" > > , OR the amount of data you lose at the end of the ten years is multiplied > by ten > - i.e. IF you know the probability of a disk crashing is "once in x > years" then you expect to lose one disks' worth of data after x years - > therefore if you have "n" disks the amount of data you lose after "x" years > is "n" disks's worth > > In the first scenario, the probability of a disk crash is OBVIOUSLY > increased if you have more disks > In the second scenario, if you amortise those "n" disks crashing over "x" > years you easily calculate that, in ONE year you will have effectively (n/x) > disks crashing - which is the same as saying you will have ONE of those "n" > disks crashing effectively every (x/n) years. You can see this is exactly > the same statement as before
I think different way. If you have 10 disks you have 10 disks which may crash. That's okay. But it has nothing to do with RAID. I say it is a nonsense to say that RAID increases possibility of disk crashes. Using many disks increases that possibility, regardless you use RAID or not. RAID is innocent!!!!! > So either way you look at it, the probability that any one of your disks > crashes is dependent upon the number of disks you have. > > As a second example, consider this. What's the probability of rolling a "6" > on a single dice? It's 1/6 , right? > Now what's the probability of rolling a "6" if you throw two dice > simultaneously? Either of the dice can turn up "6" - hey, with two dice > that means you're twice as likely to get a "6"! > - What's the probability of rolling a "6" if you throw 36 dice > simultaneously? You actually have a 600% chance of throwing a "6"! What > that means is, on AVERAGE, you will expect to throw six "6"s every time you > toss all 36 dice. > > It's not nonsense, just basic probability theory. blah blah blah You really wrote a very long mail... And the only thing which you proved is that if you have many dices you have much higher possibility to throw '6'. But you don't need RAID to do it :-) So you waste your time here. > > What data loosing are you talking about? I think hard drive failure is not a > > common problem > > (compared e.g. to strange problems of M$ Anything <enter any year here>) > > Or not? > > hard drive failure is more common than maybe you think... hard drive > manufacturers five years ago were quoting mean time between failures (MTBF) > that meant that, if you ran their drives practically non-stop for > approximately three to five years, you would expect media errors, drive > mechanical errors and bus errors to occur at least once. > I know this because, among other things, our company supplies Audio-Visual > equipment to telephone operators to record calls and supply live audio > feeds, and the disks we started supplying three-to-five years ago and now > starting to fail with alarming regularity...! > And of course, if a drive starts to fail, it's only gonna get worse ... > reformatting can only help /so much/ ... I would never use disks older than two or max. three years. At least because of insufficient capacity. So I am happy. Also I mean two real years, not two years of uninterrputed run. I usually sleep each day (or night ;-), and my computer is off then. And finally we were talking about loss of data. Not each bus error causes a loss of data. I have bus errors very often, since I bought just a stupid cheap motherboard. Loss of data which can be "repaired" by RAID is when your drive becomes physically damaged, or logically bad data is written to it. In the second case, the probability of error depends only on data size, not the number of disks. In the case the probability depends rather on number of heads. Better said if you buy now a 45MB disk, it is interanally the duplicate of three 15MB disks. So data loss probability is the same. This behaviour is caused by usual hard drive design. > > btw. Is better 133MHz bus with ATA100 drive - or - 100MHz bus with ATA66 > > :) > > (What did you expect me to say? ;OD ) Okay, I wanted to ask: Is better 100MHz bus with ATA100 drive - or - 133MHz bus with ATA66? Haven't you realized?
