Hard drive life expectancy

On Mon, 05 Jun 2017 09:59:37 -0700, Ken Blake
wrote:

On Mon, 5 Jun 2017 12:05:49 +0800, "Mr. Man-wai Chang"
wrote:

On 4/6/2017 9:14 PM, Scott wrote:
I am now wondering about the (secondary) hard drive, which I think is
probably now 10 years old (moved from an earlier computer).

Backup backup backup frequently!




Read his next paragraph, which you quoted. He says he backs up
regularly.

As OP, yes I do.

And "frequently" is not good advice for everyone. How often you should
back up depends on you and what you have on your drive.

It takes time and effort to backup, but it also takes time and effort
to recreate lost data. If you back up daily, you should never have to
recreate more than one day's worth of lost data. If weekly, there's
potentially a lot more to recreate. You should assess how much pain
and trouble you would have if you lost x days of data, and then choose
a backup frequency that doesn't involve more pain and trouble than
that you would have if you had to recreate what was lost.

That is pretty much my approach. I use Acronis each week. I also
back up one of my specific folders on to memory sticks, with two of
them which I alternate. This is on a different day, usually.

Some things (photographs, for instance) can never be recreated, and
more frequent backup may be wanted for them.

Actually, if they still in the camera the SD card there provides a
backup so I don't see any need to back up photos more frequently..

At one extreme is the professional user who would likely go out of
business if his data was lost. He probably needs to back up at least
daily. At the other extreme is the kid who doesn't use his computer
except to play games. He probably needs no backup at all, since worst
case he can easily reinstall his games.

Most of us fall somewhere between those extremes, but nobody can tell
you where you fall; you need to determine that for yourself.


I assume the risk of failure has now increased. I see 'chat' about
discs lasting four years but is this years of ownership or years of
actual use? I turn my PC off when not in use. Is there an argument
for preventive replacement or can I rely on warning signs to tell me
when the drive is on its way out (with regular back-ups of course)?

On Sun, 4 Jun 2017 10:06:08 -0400 "Mayayana"
wrote in article
also not unusual for them to die
early. In some cases there seems to be a pattern
whereby if it lives through the first few years it will
probably live many more


That's the famous "bathtub curve". Devices, components, etc., suffer from
"infant mortality" where they fail early in use but then the curve tails
off for some years before it begins to rise up again, signalling end-of-
life failures.

On Mon, 05 Jun 2017 23:07:16 -0400, Jason wrote:

On Sun, 4 Jun 2017 10:06:08 -0400 "Mayayana"
wrote in article
also not unusual for them to die early. In some cases there seems to be
a pattern whereby if it lives through the first few years it will
probably live many more


That's the famous "bathtub curve". Devices, components, etc., suffer
from "infant mortality" where they fail early in use but then the curve
tails off for some years before it begins to rise up again, signalling
end-of- life failures.

What kind of curve fit our WindowsME eMachine? Celeron 700Mhz.
The CD-ROM died early, after very little use.
The motherboard died last.
There were a couple hundred semi-functional hours between those 2 events.

Fugliest computers ever were Compaq models that looked like an Easter
Island Head going "ooooh" while receiving an enema. I gotta get one of
those.

http://www12.0zz0.com/2012/11/30/01/842198962.jpg

On Fri, 09 Jun 2017 17:31:48 +0000, Jim H
wrote:

On Sun, 4 Jun 2017 10:06:08 -0400, in ,
"Mayayana" wrote:

Lasting over 10 years is not unusual, but in test
reports I've seen it's also not unusual for them to die
early. In some cases there seems to be a pattern
whereby if it lives through the first few years it will
probably live many more.

The failure pattern for hard drives... and for that matter most
mechanical things... is that they fail early due to manufacturing
defect or last a long time after which failure rate increases. If you
plot time vs the number of failures the curve is shaped like a side
view of the inside of a bathtub. It's a well known phenomenon in the
world of Quality Assurance. Google "bathtub curve" for more detail.

I have two 300 GB SAS hard drives in my Xserves.

They are 15,000 RPM with 6.0 gig interface.
I have read that SAS hard drives are more reliable than SATA
hard drives as well as being faster.

En el artículo , Lucifer
Morningstar escribió:

I have read that SAS hard drives are more reliable than SATA
hard drives as well as being faster.

You read wrong.

--
(\_/)
(='.'=) "Between two evils, I always pick
(")_(") the one I never tried before." - Mae West

Lucifer Morningstar wrote:
On Fri, 09 Jun 2017 17:31:48 +0000, Jim H
wrote:

On Sun, 4 Jun 2017 10:06:08 -0400, in ,
"Mayayana" wrote:

Lasting over 10 years is not unusual, but in test
reports I've seen it's also not unusual for them to die
early. In some cases there seems to be a pattern
whereby if it lives through the first few years it will
probably live many more.
The failure pattern for hard drives... and for that matter most
mechanical things... is that they fail early due to manufacturing
defect or last a long time after which failure rate increases. If you
plot time vs the number of failures the curve is shaped like a side
view of the inside of a bathtub. It's a well known phenomenon in the
world of Quality Assurance. Google "bathtub curve" for more detail.

I have two 300 GB SAS hard drives in my Xserves.

They are 15,000 RPM with 6.0 gig interface.
I have read that SAS hard drives are more reliable than SATA
hard drives as well as being faster.

SAS is just a protocol. It's "like SATA, only different".

You can take a 7200RPM Seagate drive, and slap either a SATA
or SAS controller board on the back of the drive. Seagate makes
some very nice 6TB drives, 7200RPM devices, which are available
in six or more flavors. You can get the drive with a SATA controller
board, or with a SAS controller board. The addition of the SAS board
would likely raise the price a bit. And have *no* impact on
reliability of the drive.

SAS works with up to 8 meters of cabling. It has line buildout,
and likely has a single bit to flip, which declares the cable
is "long" or "short". Yours is probably set to "Short" right now.
This affects the shape of the signals on the cable, and
whether pre-emphasis is used to compensate for cable length.
At the point that SAS was introduced, SATA didn't have that.

SAS has a signal amplitude of up to 1.6Vpp, whereas SATA is
a lot less. In the article here, you can see some oscilloscope traces
for SATA (this is a re-driver IC, something not normally used on
motherboards).

https://www.maximintegrated.com/en/a...ex.mvp/id/4648

Since SATA cable length has stubbornly remained at 1 meter (SATA)
or 2 meter (ESATA), I would have to assume it doesn't have buildout,
and the SATA signal preemphasis is set for a short cable at all times.

SATA has a command queue of depth 32. SAS has a depth of 256. This
allows re-ordering of command completion, amongst other things. It
allows the drive to decide what order to service requests in.
Since SAS stands for Serially Attached SCSI, the queuing on SAS
is likely an extension of "disconnect/reselect" from parallel
SCSI days.

*******

The controller board, is a separate issue from the HDA. The
15K rotational speed, smaller platter, different stroke
size, different platter composition, operating temperature
range, power dissipation, makes the drive quite different
from a 7200RPM drive.

And that is what affects the reliability. The rotating gizmos.

The protocol, you can swap that from SAS to SATA, with hardly
any impact at all on reliability. Unless one of the controller
board chips runs too-hot, there really isn't a reason for
the board to be that much different. The chip count and
architecture are going to be the same (SOC with external
RAM and Flash, separate physical layer amplifiers or whatever).

Paul

On Sat, 10 Jun 2017 17:15:47 +0100, Mike Tomlinson
wrote:

En el artículo , Lucifer
Morningstar escribió:

I have read that SAS hard drives are more reliable than SATA
hard drives as well as being faster.

You read wrong.

Why are SAS drives used in enterprise computing?

On Sat, 10 Jun 2017 18:21:49 -0400, Paul
wrote:

Lucifer Morningstar wrote:
On Fri, 09 Jun 2017 17:31:48 +0000, Jim H
wrote:

On Sun, 4 Jun 2017 10:06:08 -0400, in ,
"Mayayana" wrote:

Lasting over 10 years is not unusual, but in test
reports I've seen it's also not unusual for them to die
early. In some cases there seems to be a pattern
whereby if it lives through the first few years it will
probably live many more.
The failure pattern for hard drives... and for that matter most
mechanical things... is that they fail early due to manufacturing
defect or last a long time after which failure rate increases. If you
plot time vs the number of failures the curve is shaped like a side
view of the inside of a bathtub. It's a well known phenomenon in the
world of Quality Assurance. Google "bathtub curve" for more detail.

I have two 300 GB SAS hard drives in my Xserves.

They are 15,000 RPM with 6.0 gig interface.
I have read that SAS hard drives are more reliable than SATA
hard drives as well as being faster.

SAS is just a protocol. It's "like SATA, only different".

You can take a 7200RPM Seagate drive, and slap either a SATA
or SAS controller board on the back of the drive. Seagate makes
some very nice 6TB drives, 7200RPM devices, which are available
in six or more flavors. You can get the drive with a SATA controller
board, or with a SAS controller board. The addition of the SAS board
would likely raise the price a bit. And have *no* impact on
reliability of the drive.

SAS works with up to 8 meters of cabling. It has line buildout,
and likely has a single bit to flip, which declares the cable
is "long" or "short". Yours is probably set to "Short" right now.
This affects the shape of the signals on the cable, and
whether pre-emphasis is used to compensate for cable length.
At the point that SAS was introduced, SATA didn't have that.

SAS has a signal amplitude of up to 1.6Vpp, whereas SATA is
a lot less. In the article here, you can see some oscilloscope traces
for SATA (this is a re-driver IC, something not normally used on
motherboards).

https://www.maximintegrated.com/en/a...ex.mvp/id/4648

Since SATA cable length has stubbornly remained at 1 meter (SATA)
or 2 meter (ESATA), I would have to assume it doesn't have buildout,
and the SATA signal preemphasis is set for a short cable at all times.

SATA has a command queue of depth 32. SAS has a depth of 256. This
allows re-ordering of command completion, amongst other things. It
allows the drive to decide what order to service requests in.
Since SAS stands for Serially Attached SCSI, the queuing on SAS
is likely an extension of "disconnect/reselect" from parallel
SCSI days.

*******

The controller board, is a separate issue from the HDA. The
15K rotational speed, smaller platter, different stroke
size, different platter composition, operating temperature
range, power dissipation, makes the drive quite different
from a 7200RPM drive.

And that is what affects the reliability. The rotating gizmos.

The protocol, you can swap that from SAS to SATA, with hardly
any impact at all on reliability. Unless one of the controller
board chips runs too-hot, there really isn't a reason for
the board to be that much different. The chip count and
architecture are going to be the same (SOC with external
RAM and Flash, separate physical layer amplifiers or whatever).

Thank you for that information.

You would be aware that SAS controllers are downward compatible
with SATA. Apple Intel Xserves can be used with SATA or SAS by
using the appropriate caddy. Does the controller recognise the type
of drive and adjust its parameters?

Paul

Lucifer Morningstar wrote:
On Sat, 10 Jun 2017 17:15:47 +0100, Mike Tomlinson
wrote:

En el artículo , Lucifer
Morningstar escribió:

I have read that SAS hard drives are more reliable than SATA
hard drives as well as being faster.
You read wrong.

Why are SAS drives used in enterprise computing?

Just remember that the "RPMs" is an independent issue
from the Protocol "SAS/SATA" used.

The 15K drives are used, for their low seek time.

An SSD has a much better seek time than that (100usec).
Even a head switch on a rotating drive (same cylinder),
takes 1000usec. Hard drives can never catch up to SSDs,
no matter how fast they spin, because of limitations like that.

There are ways to make hard drives better, but the history
of hard drives is, "nobody cared". You will notice a
fixation with just one way of doing things, when it comes
to the design of hard drives.

Paul

On Sat, 10 Jun 2017 20:46:51 -0400, Paul
wrote:

Lucifer Morningstar wrote:
On Sat, 10 Jun 2017 17:15:47 +0100, Mike Tomlinson
wrote:

En el artículo , Lucifer
Morningstar escribió:

I have read that SAS hard drives are more reliable than SATA
hard drives as well as being faster.
You read wrong.

Why are SAS drives used in enterprise computing?

Just remember that the "RPMs" is an independent issue
from the Protocol "SAS/SATA" used.

The 15K drives are used, for their low seek time.

An SSD has a much better seek time than that (100usec).
Even a head switch on a rotating drive (same cylinder),
takes 1000usec. Hard drives can never catch up to SSDs,
no matter how fast they spin, because of limitations like that.

SSDs are not used in enterprise computing because they are
unreliable under heavy usage.

There are ways to make hard drives better, but the history
of hard drives is, "nobody cared". You will notice a
fixation with just one way of doing things, when it comes
to the design of hard drives.

It was thought that rotating disks would be long gone by now
but they persist due to high cost of SSDs and greater reliability
of mechanical hard drives.

Paul

Lucifer Morningstar wrote:


Thank you for that information.

You would be aware that SAS controllers are downward compatible
with SATA. Apple Intel Xserves can be used with SATA or SAS by
using the appropriate caddy. Does the controller recognise the type
of drive and adjust its parameters?

I read reviews on this. I have yet to see a happy
customer, who put a SATA drive on a SAS controller. The speed
is all over the place. I have no idea why people experience
such poor results, when the standards have gone to all that
trouble to make it work with both.

So yes, you can connect a SATA drive to a SAS controller.
Will you be happy with the results ? I'm still waiting
for a review that said "it just worked".

And I've had behaviors like that on SATA, but for a suspected
reason. I had a chipset here, where if you plug in a WinTV
card with BT878 onto the PCI bus, it causes the SATA controller
to "go nuts". You get data transfer rates of 30MB/sec right
after the OS boots (when normally it would be 135MB/sec). And
after some time, it drops down to 20MB/sec. This is caused
by the *BIOS* using a patch for a perceived chipset problem
on the PCI bus. A problem that's probably been around
for ten plus years, and never fixed by VIA. But the people
with bad SAS board behaviors, it's not going to be a situation
like that. The motherboard that was doing that, I "retired" it.
It was otherwise a very nice motherboard. Once the WinTV card
is unplugged, the SATA speed jumps back up to "normal", so there
is no damage.

Paul

On Sun, 11 Jun 2017 05:00:11 -0400, Paul
wrote:

Lucifer Morningstar wrote:


Thank you for that information.

You would be aware that SAS controllers are downward compatible
with SATA. Apple Intel Xserves can be used with SATA or SAS by
using the appropriate caddy. Does the controller recognise the type
of drive and adjust its parameters?

I read reviews on this. I have yet to see a happy
customer, who put a SATA drive on a SAS controller. The speed
is all over the place. I have no idea why people experience
such poor results, when the standards have gone to all that
trouble to make it work with both.

AFAIK most people who used Xserves used SATA drives.
I have not been aware of any speed problems in my
use of the Xserves in my hobby.
Sometimes the Xserve will not recognise a SATA drive right
away. I wanted to copy some files from a 320 GB USB drive
but the Xserve did not recognise the drive. I have 1 TB
WD drive taken from an external drive box where the USB
interface failed. The Xserve did not see it at all until I put
a link in the back to reduce the interface speed.

So yes, you can connect a SATA drive to a SAS controller.
Will you be happy with the results ? I'm still waiting
for a review that said "it just worked".

And I've had behaviors like that on SATA, but for a suspected
reason. I had a chipset here, where if you plug in a WinTV
card with BT878 onto the PCI bus, it causes the SATA controller
to "go nuts". You get data transfer rates of 30MB/sec right
after the OS boots (when normally it would be 135MB/sec). And
after some time, it drops down to 20MB/sec. This is caused
by the *BIOS* using a patch for a perceived chipset problem
on the PCI bus. A problem that's probably been around
for ten plus years, and never fixed by VIA. But the people
with bad SAS board behaviors, it's not going to be a situation
like that. The motherboard that was doing that, I "retired" it.
It was otherwise a very nice motherboard. Once the WinTV card
is unplugged, the SATA speed jumps back up to "normal", so there
is no damage.

I have four Intel Xserves on my home network but only two
SAS hard drives. I use the SAS drives as boot and system disks
on two of the Xserves. Is that the best use of them?



Paul

Lucifer Morningstar wrote:


I have four Intel Xserves on my home network but only two
SAS hard drives. I use the SAS drives as boot and system disks
on two of the Xserves. Is that the best use of them?

Which are you worried about most ?

The 15 seconds it takes to boot in the morning ?

Or the extra 60% user load it can handle during
the day ? If you have data services to offer some
group, then a faster data disc means more IOPS.

If the discs aren't big enough to offer effective
data services, then obviously they won't be data
drives. And then you can use them as boot drives
if you want.

And they're not "SAS" drives, as much as they're "15K"
drives. That's why you're using them, is for the lower
seek time offered by a 15K drive.

Paul

On Wed, 14 Jun 2017 01:31:38 -0400, Paul
wrote:

Lucifer Morningstar wrote:


I have four Intel Xserves on my home network but only two
SAS hard drives. I use the SAS drives as boot and system disks
on two of the Xserves. Is that the best use of them?

Which are you worried about most ?

I should explain. It's just a hobby and I am the only user.
I do use the servers for backup.
I picked up the two 300 GB 5.25" SAS drives from an online
auction and the last Xserve I bought came with three empty
SAS caddys.

The 15 seconds it takes to boot in the morning ?

That was what I was thinking but a lot of the startup time
is taken by the testing the server does before accessing
the boot disk.

Or the extra 60% user load it can handle during
the day ? If you have data services to offer some
group, then a faster data disc means more IOPS.

Presumably the transfer rate is limited by my gigabit network so
probably no difference.

If the discs aren't big enough to offer effective
data services, then obviously they won't be data
drives. And then you can use them as boot drives
if you want.

Thanks.

And they're not "SAS" drives, as much as they're "15K"
drives. That's why you're using them, is for the lower
seek time offered by a 15K drive.

I also have a Windows server with two 2.5" 146 GB SAS
drives and a SPARC machine which is running Solaris 10
and which has four 2.5" 73 GB SAS drives.

Paul

On Sun, 11 Jun 2017 05:00:11 -0400, Paul wrote:

Lucifer Morningstar wrote:


Thank you for that information.

You would be aware that SAS controllers are downward compatible
with SATA. Apple Intel Xserves can be used with SATA or SAS by
using the appropriate caddy. Does the controller recognise the type
of drive and adjust its parameters?

I read reviews on this. I have yet to see a happy
customer, who put a SATA drive on a SAS controller.

*Now* you tell me. I've been running SATA drives off of a pair of SAS
controllers since early 2009 without any issues. I'm aware of at least a
few hundred others who are doing the same thing, according to the forums
at www.avsforum.com.

The speed
is all over the place. I have no idea why people experience
such poor results, when the standards have gone to all that
trouble to make it work with both.

I use 5400 RPM drives, so the speed may be limited by that. I get the
speed that I expect, which is very much what I get from the same model
of drive when it's connected directly to a mobo SATA connector.

So yes, you can connect a SATA drive to a SAS controller.
Will you be happy with the results ? I'm still waiting
for a review that said "it just worked".

Two controllers, a total of 16 drives, and it "just worked".

Each controller has two SAS ports. Each of those ports uses a "SAS to
SATA" breakout cable, giving me 4 SATA ports per SAS port. The cards use
the Marvell 88SE64xx chipset.

Now that you've told me to expect issues, I suppose it'll all crash and
burn in short order. It was perfectly fine when I didn't know.