Prev: Microsoft Fix It
Next: Computer crash
From: John on 17 Jun 2010 19:14
I'm confused about terminology used for SSD devices and would
appreciate clearing up what I'm overlooking.

If a SSD has no moving parts, then why does the literature reference
head seek times and estimate the number of head seeks before the device
would fail. It seems contradictory to me because I keep thinking of
head seeks as being an actual movement within drive yet the ads lead me
to expect that the drive has no moving devices inside it.

I would have thought failure would be due to something of a
non-mechanical nature.

When the literature talks about the number of seeks that an SSD can
peform before failing, what does that correspond to (roughly) in actual
time? I'm used to thinking of hard disk drives as lasting around five
years or more based upon my usage. I'm trying to develop some picture
of what would be the equivalent for an SSD device, especially since I'm
starting to see recommendations to use an SSD for the boot disk in a
pc. I have no experience with them to figure out how long they might
last and what to listen for to be aware if one is starting to fail.

Thanks,

John
From: Paul on 17 Jun 2010 23:16
John wrote:
> I'm confused about terminology used for SSD devices and would
> appreciate clearing up what I'm overlooking.
>
> If a SSD has no moving parts, then why does the literature reference
> head seek times and estimate the number of head seeks before the device
> would fail. It seems contradictory to me because I keep thinking of
> head seeks as being an actual movement within drive yet the ads lead me
> to expect that the drive has no moving devices inside it.
>
> I would have thought failure would be due to something of a
> non-mechanical nature.
>
> When the literature talks about the number of seeks that an SSD can
> peform before failing, what does that correspond to (roughly) in actual
> time? I'm used to thinking of hard disk drives as lasting around five
> years or more based upon my usage. I'm trying to develop some picture
> of what would be the equivalent for an SSD device, especially since I'm
> starting to see recommendations to use an SSD for the boot disk in a
> pc. I have no experience with them to figure out how long they might
> last and what to listen for to be aware if one is starting to fail.
>
> Thanks,
>
> John

As far as I know, flash has limited program/erase cycles, but for
reading, you can do as much as you want.

NAND flash is block oriented, but that is generally hidden from the user
(changing a 512 byte sector means erasing and writing a 128KB block).
Wear leveling, is intended to spread the limited program/erase cycles
over the flash device. You can take the capacity (32GB) multiply by
the program cycles (10K for MLC type flash perhaps) and get some idea
of the amount of data that can be written (32GB*10K). Some manufacturers
might state that as "writing at 40MB/sec for 5 years", as a way of stating
what they think the wearout will be. Some of the flash device capacity is
held in reserve, for more efficient operation or as sparing for bad blocks.

In terms of failure modes, you can see some fail in this NASA testing.
But they're using irradiation for testing, which is an issue in space.
Irradiation on earth could still be an issue, as we're still exposed
to cosmic radiation.

http://radhome.gsfc.nasa.gov/radhome/papers/tns2006_Oldham.pdf

Do you have a link to the document in question ? Perhaps it
contains some justification for using an inappropriate analogy
about "heads n' seeks".

Paul
From: John on 18 Jun 2010 10:12
Paul wrote:

> John wrote:
> > I'm confused about terminology used for SSD devices and would
> > appreciate clearing up what I'm overlooking.
> >
> > If a SSD has no moving parts, then why does the literature reference
> > head seek times and estimate the number of head seeks before the
> > device would fail. It seems contradictory to me because I keep
> > thinking of head seeks as being an actual movement within drive yet
> > the ads lead me to expect that the drive has no moving devices
> > inside it.
> >
> > I would have thought failure would be due to something of a
> > non-mechanical nature.
> >
> > When the literature talks about the number of seeks that an SSD can
> > peform before failing, what does that correspond to (roughly) in
> > actual time? I'm used to thinking of hard disk drives as lasting
> > around five years or more based upon my usage. I'm trying to
> > develop some picture of what would be the equivalent for an SSD
> > device, especially since I'm starting to see recommendations to use
> > an SSD for the boot disk in a pc. I have no experience with them
> > to figure out how long they might last and what to listen for to be
> > aware if one is starting to fail.
> >
> > Thanks,
> >
> > John
>
> As far as I know, flash has limited program/erase cycles, but for
> reading, you can do as much as you want.
>
> NAND flash is block oriented, but that is generally hidden from the
> user (changing a 512 byte sector means erasing and writing a 128KB
> block). Wear leveling, is intended to spread the limited
> program/erase cycles over the flash device. You can take the capacity
> (32GB) multiply by the program cycles (10K for MLC type flash
> perhaps) and get some idea of the amount of data that can be written
> (32GB*10K). Some manufacturers might state that as "writing at
> 40MB/sec for 5 years", as a way of stating what they think the
> wearout will be. Some of the flash device capacity is held in
> reserve, for more efficient operation or as sparing for bad blocks.
>
> In terms of failure modes, you can see some fail in this NASA testing.
> But they're using irradiation for testing, which is an issue in space.
> Irradiation on earth could still be an issue, as we're still exposed
> to cosmic radiation.
>
> http://radhome.gsfc.nasa.gov/radhome/papers/tns2006_Oldham.pdf
>
> Do you have a link to the document in question ? Perhaps it
> contains some justification for using an inappropriate analogy
> about "heads n' seeks".
>
> Paul

Thanks Paul.

After reading the article you cited and going back to articles on SSD
and flash memory at Wikipedia, it's obvious to me that I was applying
terminology and mental images appropriate to what I had learned about
hard drives from years past to the current SSD drives, apples and
oranges as it were.

John
 | 
Pages: 1
Prev: Microsoft Fix It
Next: Computer crash