View Single Post
  #49 (permalink)  
Old May 27th 15, 02:06 AM posted to uk.comp.os.linux,uk.rec.audio
Johnny B Good
external usenet poster
 
Posts: 65
Default Flac and Audio CD Health Checks

On Sat, 23 May 2015 21:03:19 +0100, John Williamson wrote:

On 23/05/2015 12:55, Jim Lesurf wrote:
In article , Tim Watts
wrote:
I would not worry about SSD wear - I have been using them in laptops
for a while - my current one is over a year old and shows 100% on the
Wear_Leveling_Count (that's good, 0% is "dead").


I'm not particularly worried about SSD wear as I've also (so far) used
some for years. But some of the people who use my program may be using
conventional spinning rust. And there *are* wear mechanisms for SSD
according to reports I've read. Just that in practice I (and many
others) don't seem to have run into problems with it.

From what I've read, early flash memory cells had a limited life,
measured in thousands of writes, with unlimited reads being safe, while
new types have write lives in the millions of cycles due to improvements
in manufacturing techniques as the technology has matured. New ones also
have more redundancy built in, so the controller can transparently
replace failing cells with unused ones until the spares have all been
used.


You basically have that arse about face. The very early flash devices
had erase/write cycles initially measured in tens of thousands, rising to
hundreds of thousands culminating in millions before the controllers
gained wear levelling capabilities that allowed smaller scale nand cells
with reducing erase/write cycle limits that allowed larger nand cell
counts on the die.

The early SSDs with proper wear levelling used slc nand with e/w cycle
ratings of 30000 or so which, due to reducing cell dimensions and mlc tlc
have e/w cycle lifetimes as low as a mere 3000 or even less. The overall
endurance ratings of modern SSDs are the result of much larger
capacities, some over provisioning (7 to 15% afaicr) and yet more
sophisticated wear levelling controller technology.

The real key to being able to get acceptable life out of a modern SSD
lies with the sheer capacity increases which now require a petabyte or
more's worth of write activity to wear out a 512GB SSD that's using nand
with only a nominal 1500 to 3000 e/w cycle rating thanks, essentially due
to the clever way the controller can spread the wear amongst the nand
cells in the whole array.

A quick 'n' dirty calculation reveals that writing as much as 100GB a
day to a 512GB SSD using nand cells rated for 1500 cycles with the
benefit of wear levelling should last for just over 20 years of 7 days a
week usage. You'd have to be writing/erasing a good 400GB a day to reduce
the service life of such an SSD to a mere 5 years.

In practice, the vast majority of home PC users are unlikely to subject
their systems to much more than 1 to 10 GBs per day. Even people like me
recording several gigabyte's worth of freeview media files a day are
unlikely to top 30GB a day with SD content (or 50GB a day with HD
content).

By the time you've worn out a 500GB SSD, the time for an upgrade will be
long overdue and you'll likely be looking at a 5TB unit with nand cells
rated for a mere 300 cycles or less but, due to the higher capacity, will
last at least the next 5 years or so despite a guesstimated quadrupling
in data throughput.

Ultimately, we may see petabyte sized SSDs with nand cell cycle limits
of just ten cycles which will last 5 years simply because it will take 6
months of 24/7 operating time just to completely write a single set of
data to completely exhaust the SSD's capacity!

Indeed, the ultimate may be to do away with the erase cycle altogether
once capacities of data storage devices start being measured in dozens of
petabytes. Clever controller technologies will probably start using
techniques whereby older data can be recycled to simulate the storage of
new data by clever remapping and 'diffing'. Obviously this will require a
relatively small proportion of high cycle endurance nand for the
controller's own use but the vast majority could consist of worm type
memory cells with the much greater density worm chips making up the bulk
of the storage device.

In the end, your only means of EOL 'secure erasure' for those petabyte
sized SSDs may simply be an electric furnace with an SSD sized slot
through to which to feed your redundant storage devices. Mind you, it's
very likely by then that HMG will have made GCHQ provide a free backup
service as part of the covenant of the universal surveillance pact with
the electorate. :-)

--
Johnny B Good