AHCI and IDE

[Flasherly]

But for those of you who just want to know the conclusions, here are the really interesting points of the article:

Interesting.

Say I've never had hard drives fail, at least good hard drives. Say
again I'm able to find or at least know a good from a bad drive. Both
statements could be saying the one thing I -do not- want to say about
hard drives: My experience with drives which have failed, and whether
there was anything preventative, on my part, I could have done better
to prolong their longevity.

Heat and usage are the only two subjective factors Google alludes to
in otherwise a statistical or objective skew of livespan. On that
skew the correlation is three factors: 1) the chance of premature
failure, the chance of long-term failure, and the chance either is
predictable.

Elsewhere. Newegg. What I'm seeing over a widest distribution of
responses to having purchased a particular drive model is that 1 in 4
are at extremes, that 1 is totally dissatisfied, whereas another 3,
for the most, are variously satisfied.

There's only one exception, a Seagate 500G model priced for a 5-yr
warranty, at entry 2T drives in other brand makes, with a stipulation
Seagate is touting for higher specifications and quality in
specifically addressing longevity. (The number of response returns I
recall high, although perhaps not as large as some of the other
models.)

The salient factor, however, is one in four, 25% is quite a larger
discrepancy from Google's potential 9%. A hard drive either works or
not. Nobody is really complaining about noise, or 5400 green drive
wouldn't match higher performance 10K or 7200 black. In order to give
it a 1 or zero, the drive was, for them, is "broken" upon delivery.
Not necessarily an impression every IT specialist runs back to provide
upon installing cases of 50 drives, although for a "figure" among
abstracts, one neither wholly dismissible.

 

[Loren Pechtel]

I've sent back everything from the 80GB range up to 1TB. I don't think
I've sent a 2TB back, and certainly not a 3TB, but I might simply not
have enough of them in service yet.

I've sent 2's back. Never a 3 but like you the quantity is low.

 

[Loren Pechtel]

Here's the SSDlife page where they explain.
http://ssd-life.com/eng/how.html

Argue with them if you like. My SSD is running nicely and fully imaged.
If it packs up next week I'll look for an explanation but I won't be
devastated.

I'm looking at the output of their own program. A drive used to host
multiple virtual machines:

Used: 1 year, 8 months. Life remaining 99%. That would give a life
expectancy of at least a century.

Estimated lifetime: 8 years, 2 months.

In other words, it's capping it at just over 10 years.

 

[Robin Bignall]

I'm looking at the output of their own program. A drive used to host
multiple virtual machines:

Used: 1 year, 8 months. Life remaining 99%. That would give a life
expectancy of at least a century.

Estimated lifetime: 8 years, 2 months.

In other words, it's capping it at just over 10 years.

I don't know what your point is. None of my HDDs have been around for
10 years: the oldest (usually the smallest or slowest) gets replaced
every 3/4 years when I build a new machine. So, I expect, will the SSD
if I live that long.

 

[RayLopez99]

I don't know what your point is. None of my HDDs have been around for

10 years: the oldest (usually the smallest or slowest) gets replaced

every 3/4 years when I build a new machine. So, I expect, will the SSD

if I live that long.

I think Loren is onto something. Thanks for the link but notice this: fromthe link that you provided, they refer to Intel's owns statements about their X25-M SSD, which AnandTech analyzed in 2008 here: http://www.anandtech.com/show/2614/4

And the conclusion: five years under heavy usage.

"
Note: by the way, some manufacturers give the total amount of data written to the drive as one of the drive lifetime indicators. For example, Intel guarantees that the total of about 37 TB of data will be written to X25-M drives (20 GB per day for 5 years: “The drive will have a minimum of 5 yearsof useful life under typical client workloads with up to 20 GB host writesper day.”).
"

So that's your answer: Intel will guarantee this SSD drive for three years, and under heavy usage you'll get five years, and lighter usage more years..

Not bad. Not bad at all.

RL

 

[Robin Bignall]

I think Loren is onto something. Thanks for the link but notice this: from the link that you provided, they refer to Intel's owns statements about their X25-M SSD, which AnandTech analyzed in 2008 here: http://www.anandtech.com/show/2614/4

And the conclusion: five years under heavy usage.

"
Note: by the way, some manufacturers give the total amount of data written to the drive as one of the drive lifetime indicators. For example, Intel guarantees that the total of about 37 TB of data will be written to X25-M drives (20 GB per day for 5 years: “The drive will have a minimum of 5 years of useful life under typical client workloads with up to 20 GB host writes per day.â€).
"

So that's your answer: Intel will guarantee this SSD drive for three years, and under heavy usage you'll get five years, and lighter usage more years.

Not bad. Not bad at all.

Right! (I replied to your email.)

 

[Loren Pechtel]

I don't know what your point is. None of my HDDs have been around for
10 years: the oldest (usually the smallest or slowest) gets replaced
every 3/4 years when I build a new machine. So, I expect, will the SSD
if I live that long.

The point is SSD life is not producing a life estimate based on wear.
That's why his end of life date hasn't changed.

 

[Yousuf Khan]

Incidentally, in Sentinel, if I click on the SSD and Information, down
the list a long way is something like SSD Features. The final entry
says "OS TRIM Function...... Supported, Disabled"

My best guess at what this means, since TRIM IS enabled in Win7, is ...
WTF does it mean?

I don't know why yours says that, but mine shows "Supported, Enabled".
Are you still seeing this message now, or is this an old message? Here's
an FAQ from Corsair about how to enable TRIM support. Even if you don't
have a Corsair SSD, it's still relevant.

How to Enable/Check TRIM/AHCI in Windows 7 - The Corsair Support Forums
http://forum.corsair.com/forums/showthread.php?t=86403

Yousuf Khan

 

[Yousuf Khan]

Most SSDs have some sort of garbage collection that partially negates
the need for TRIM. It doesn't really solve the problem as well as having
the OS say "This block isn't needed anymore", but it helps.

Yes, garbage collection still occurs without TRIM, however TRIM just
tells the drive that the OS is not very busy right now, so it's safe to
run the background garbage collection procedure now. That is, it'll run
the GC without any worry of interfering with processes running on the
machine.

I'm not fully clear about how exactly drives do their own garbage
collection either, as far as I can tell they either need a basic
understanding of filesystem data structures (which won't exist in a
RAID-0/5 environment), or they would need to rely on the OS occasionally
wiping the drive with NULLs (which is more or less how, in the virtual
machine world, a guest OS can indicate that certain blocks aren't needed
anymore)

When a block of data needs to be rewritten or simply erased, the OS
sends a write command to that block. But that specific block doesn't
actually get written to, instead the drive itself marks that block as
dirty and sends the OS a totally different block that is marked as
unwritten. The OS has no idea that it's being presented a totally
different block, therefore the OS has no internal knowledge of the
organization of the flash memory. The drive then erases a whole group of
blocks in a single erase operation in the background which can then be
presented as unwritten blocks to the OS, the next time it needs them.

Strangely enough, in flash memory, erased blocks aren't filled with
zeros, but filled with ones! The write operation to a flash block goes
around changing ones to zeros, and/or leaving ones in place. It sounds a
little backward, but it accomplishes the same thing.

Yousuf Khan

 

[Robin Bignall]

I don't know why yours says that, but mine shows "Supported, Enabled".
Are you still seeing this message now, or is this an old message? Here's
an FAQ from Corsair about how to enable TRIM support. Even if you don't
have a Corsair SSD, it's still relevant.

An old message. It's ticked now.

 

[DevilsPGD]

In the last episode of <[email protected]>,

I've sent 2's back. Never a 3 but like you the quantity is low.

I've only got 3 2TBs online and don't plan on getting more, so I
wouldn't expect to see any real issues. We've got 6 3TB drives online
now, and will be standardizing on 3TB for the near future, so I expect
we'll start to see failures over time, just like every other size,
brand, speed and type that we've used. Such is life.

Luckily storage is so cheap that we can afford substantial redundancy.
Right now for our primary data, it's stored either RAID-1 or RAID-5
(mostly RAID-1, only one -5 array left that I'm aware of) on each
server, with pairs of servers redundant to each other. One of those
pairs is responsible for maintaining offsite snapshots on a regular
basis, each of which is stored in at least 3 datacenters, with
redundancy within each datacenter too, all for prices that are low
enough to be a rounding error when compared to our total IT budget.

Unimportant data is still stored with at least one locally redundant
copy, once on each server, primarily so that when a server goes offline
we don't lose access to the data, but we don't store it off-site, mainly
due to the time to upload vs the value of the data (and/or the costs of
rebuilding)

 

[Loren Pechtel]

Luckily storage is so cheap that we can afford substantial redundancy.
Right now for our primary data, it's stored either RAID-1 or RAID-5
(mostly RAID-1, only one -5 array left that I'm aware of) on each
server, with pairs of servers redundant to each other. One of those
pairs is responsible for maintaining offsite snapshots on a regular
basis, each of which is stored in at least 3 datacenters, with
redundancy within each datacenter too, all for prices that are low
enough to be a rounding error when compared to our total IT budget.

Yup. The only stuff I have that's not RAID is the SSD.