What am I doing wrong ??? Or is Adaptec 21610SA just a crappy RAID card ?

Arno Wagner wrote in
:

[snip]
It is quite clear which one you want in a high-priced
server.

ATA for a CDROM and Fibre Channel for the rest, please.


--
/Jesper Monsted

looks like MTBF about 100,000 (assuming 180 drives get replaced). No good.

"Jesper Monsted" wrote in message
.163...
"Maxim S. Shatskih" wrote in news:coq0t2$rs7$1
@gavrilo.mtu.ru:

No, this only means that each year 1 of 136 disks will fail

Your calculations just don't match the real world, but what the hell

Out of 2200 or so 146GB FC drives, we replace 2-5 every week. This is
quite
a bit more than your one-in-136 a year.

--
/Jesper Monsted

Jesper Monsted wrote:

"Maxim S. Shatskih" wrote in news:coq0t2$rs7$1
@gavrilo.mtu.ru:

No, this only means that each year 1 of 136 disks will fail

Your calculations just don't match the real world, but what the hell

Out of 2200 or so 146GB FC drives, we replace 2-5 every week. This is
quite a bit more than your one-in-136 a year.

(a) what is the MTBF on those drives
(b) if the failure rate is significantly greather than the rated MTBF would
suggest then have you tried to find out what is killing them?


--
--John
Reply to jclarke at ae tee tee global dot net
(was jclarke at eye bee em dot net)

In article ,
"J. Clarke" wrote:

flux wrote:

This supports my assertion that gigabit is relatively recent.

How? By that reasoning keyboards are "relatively recent" because Dell's
laptops have them.

I think Dell always has shipped their laptops with keyboards :-) But
when did they start shipping with gigabit? Recently, no?

Well, dude, you design one and see how you make out.

They need to be designed?

Your calculations just don't match the real world, but what the hell

Out of 2200 or so 146GB FC drives, we replace 2-5 every week. This is
quite
a bit more than your one-in-136 a year.

That could be because:
1. Your hard drives work harder
2. Power on Hours per year are higher
3. Ambient temperature is higher
4. Drives are getting older
than specified for MTBF measurement.

Peter wrote:

Your calculations just don't match the real world, but what the hell

Out of 2200 or so 146GB FC drives, we replace 2-5 every week. This is
quite
a bit more than your one-in-136 a year.

That could be because:
1. Your hard drives work harder
2. Power on Hours per year are higher

One would hope that for enterprise-quality drives the design hours per year
would be 8766.

3. Ambient temperature is higher
4. Drives are getting older
than specified for MTBF measurement.

--
--John
Reply to jclarke at ae tee tee global dot net
(was jclarke at eye bee em dot net)

"J. Clarke" wrote in
:

Jesper Monsted wrote:

"Maxim S. Shatskih" wrote in
news:coq0t2$rs7$1 @gavrilo.mtu.ru:

No, this only means that each year 1 of 136 disks will fail

Your calculations just don't match the real world, but what the hell


Out of 2200 or so 146GB FC drives, we replace 2-5 every week. This is
quite a bit more than your one-in-136 a year.

(a) what is the MTBF on those drives

Not quite sure, since it's a mix of whatever EMC had on the shelf at the
time. There are quite a few seagates (rated at 1,400,000 hours and
1,200,000 hours, depending on the model) and Ultrastars (which i can't find
the MTBF for, but assume it's about the same).

(b) if the failure rate is significantly greather than the rated MTBF
would suggest then have you tried to find out what is killing them?

Firmware. They are being replaced as soon as anything looks like it's going
to fail. Soft errors, S.M.A.R.T. readouts etc is all taken into
consideration and drives preemptively replaced.

They still managed to fail two in the same raidset, b0rking 18 TB of
datawarehouse, though.

--
/Jesper Monsted

"Peter" wrote in news:31i5k1F3asuueU1
@individual.net:

Your calculations just don't match the real world, but what the hell

Out of 2200 or so 146GB FC drives, we replace 2-5 every week. This is
quite
a bit more than your one-in-136 a year.

That could be because:
1. Your hard drives work harder

Sitting in a enterprise class storage system, they might, but that
shouldn't fail them that often.

2. Power on Hours per year are higher

24/7, just like they were designed for (i hope... who ever turns off
servers?)

3. Ambient temperature is higher

Properly cooled datacenter at about 22 degrees C. Exhaust air from the box
isn't even warm to the touch.

4. Drives are getting older
than specified for MTBF measurement.

About a year and a half now, but this has been the same since we turned on
the things.

As i noted in a different post, we (or rather, EMC) replace them as soon as
they act funny in any way, which probably means our definition of
"failure" is a bit different than the disc manufacturers...

I was told EMC scrapped a 100k unit shipment of ultrastars at one point -
there's a bad RMA for some support droid at IBM

--
/Jesper Monsted

"Peter" wrote in message
Your calculations just don't match the real world, but what the hell

Out of 2200 or so 146GB FC drives, we replace 2-5 every week. This is
quite a bit more than your one-in-136 a year.

That could be because:
1. Your hard drives work harder

They are FC drives.

2. Power on Hours per year are higher

They are FC drives.

3. Ambient temperature is higher

They are FC drives.

4. Drives are getting older

Or at 146GB are still very young.

than specified for MTBF measurement.

146GB drives, older than 5 years? .

In article ,
Peter wrote:
Your calculations just don't match the real world, but what the hell

Out of 2200 or so 146GB FC drives, we replace 2-5 every week. This is
quite
a bit more than your one-in-136 a year.

That could be because:
1. Your hard drives work harder
2. Power on Hours per year are higher
3. Ambient temperature is higher
4. Drives are getting older
than specified for MTBF measurement.

Or it could be because unless you pay for an application-specific MTBF
guarantee (which you will probably be contractually bound to not
disclose), what you're working with is a measurement so fundamentally
tied to marketing purposes that it's basically not useful for any
technical purpose at all.

You could reasonably think of those 1,200,000 hour MTBF numbers as
being generated like this:

"We want to quote a million-hour MTBF. How many drives to we have to
run for a month to quote that? We only have a month left in the
development cycle before we have to have the marketing materials ready
for product announcement."

"Hm. I see it's a little under 2000 drives. Well, let's pull 2,000 of the
first production run (that pass the initial QA test, which will exclude DOA
units and units with the kind of obvious mechanical problems that will
kill them early; this sort of QA is often _reduced_ for later production
runs) and put them in a room for 1000 hours. If only one or two fail,
we'll claim a 1,000,000 hour MTBF and nobody will be able to sue us."

Of course very few units actually fail in the first month of use, in
the "enterprise drive" space where even moderately rigorous QA is done
on the production line. The whole key to the fraud is in manipulating
the total length of the MTBF test so that it *never hits* the actual
point in time when wear on any component might cause that component to
have a significant likelihood of causing a unit failure.

The moral of the story is that if you run your disk drives for a month
and then throw them away, you can feel reasonably confident in trusting
manufacturer MTBF numbers. Otherwise, though... caveat emptor.

--
Thor Lancelot Simon
Am I politic? Am I subtle? Am I a Machiavel?
-William Shakespeare