# To RAID or not to RAID? - Page 2

•  Subject
• Author
• Posted on

## Re: To RAID or not to RAID?

But not when the array was originally constructed, and all the drives were
new?

--
Ed Light

Smiley :-/
MS Smiley :-\

Send spam to the FTC at
uce@ftc.gov
Thanks, robots.

## Re: To RAID or not to RAID?

Ed Light wrote:

MTBF = ***MEAN*** time between failure.....  A new drive could die
tomorrow or could last for 20 years.

## Re: To RAID or not to RAID?

Ed Light wrote:

He said on *average* a harddrive would fail every week, basically at the
end of the 5 years most or all of the drives would (probably) be dead.
(The MTBF used here is very low compared to harddrives of today at 43,800).

However you are correct that at the start of the array the likely hood
of each indivual drive failing is slim. BUT The probablity of any drive
in the array failing gets higher.

ie. Lets say on day 1 the chance of a drive failing is 99.9%
Probability of a particular drive failing is 0.1% (1 in 1000 chance)
Probability of a random drive failing in the 260 2.6 % (1 in 38 chance)

Basically with a mirror you are doubling the chances of failure, if you
regulary back up, get good disks, and renew them regulary you may well
be safe, but you may well not be. Depends on how big a gambler you are.

Ash.

## Re: To RAID or not to RAID?

andrew wrote:
[...]

I should point out that MTBF figures for hard drives are not quite what you
might expect at first glance. From below I see you have the right idea -
this post is for other people :) Your average new SCSI drive has a MTBF of >
10^6 hours; the current Seagate Cheetahs have a 1400000 hour MTBF IIRC. Note
that 1.4 million hours is approximately 160 years! This doesn't mean at all
that you can run a batch of HDDs for a few hundred years until they all
fail, then calculate the average to be ~160 years. What is actually means is
that if you replace a drive at regular intervals (usually the warranty
length of the drive; 5 years for a standard SCSI drive) then you will only
have a drive fail on you (on average) every 160 years. Of course, this says
nothing about the failure rate of a single drive after it passes the 5 year
mark.

[...]

--
Michael Brown
www.emboss.co.nz : OOS/RSI software and more :)
Add michael@ to emboss.co.nz ---+--- My inbox is always open

## Re: To RAID or not to RAID?

Your math is incorrect here.  The MEAN time is 5 years, so on average,
roughly half the drives will fail in less than 5 years, and half in more
than 5 years (technically, that would be the MEDIAN, but it's pretty close).
Your example has all of the drives failing in less than 5 years, so your
mean is around 2.5 years.

Bill G

## Re: To RAID or not to RAID?

Ed Light wrote:

The chance of a used drive going bad is 100%. The question of when they

Ash.

## Re: To RAID or not to RAID?

I want to thank everyone who took the time to create a reply in this
thread, especially Derek and Ed for their links as well as Mike B. for
his in-depth replies.  The storagereview.com link from Derek convinces
me I should go with a RAID config for performance, however the price of
Matxor and WD 250 gig drives right now is very tempting to me, like
buying three 80g drives for the price of two.  As far as the risk of
losing data, well you're supposed to burn your critical data to CD or
DVD, right?  But I thought I'd simply clone any data on SATA to IDE so
at least I'll have it around.  I hope IDE will remain available and
cheap for a long, long time.  Again fellahs, thaks for all your help.