Feeds

back to article 'Wear levelling' - a bedroom aid for multi-layer cell Flash

Multi-level cell NAND flash doesn't last as long as single-level cell flash but there are ways to increase its life, under the generic heading of wear-levelling. NAND flash cells have a finite life, in that they only support a specific number of writes before failing to return valid data from a read request. It's necessary to …

COMMENTS

This topic is closed for new posts.

SSD Failure

I would be interested to hear from anyone who has been using an SSD nearing the end of its life as a Windows boot drive. Does the drive start throwing read errors much like a traditional spinning disk or do you see the capacity slowly decreasing as the controller marks more and more of the drive as unusable?

2
0
Bronze badge

It will be difficult to find anyone.

My desktop 160GB Intel SSD has 5409 power on hours and 2.9TB of writes. In other words after more than 6 months of powered usage its FLASH has had about 18 write cycles.

Amplify that by 2.7 or something and I will still be dead long before it wears out. The SMART media wearout indicator is 99 of 100 (and I'm not sure it wasn't 99 to start with).

It is hard to imagine a desktop usage pattern where a half decent SSD wears out before it fails for other reasons or becomes obsolete.

0
0
Bronze badge
Unhappy

On my case...

My laptop one day started to behave oddly, I rebooted it, and seemed to boot just to crash half way... fsck later revealed huge amounts of filesystem errors... format did no good.

It wasn't nice or pleasant.

0
0
Unhappy

Where do we keep data now?

Modern hard drives, with huge areal densities are flakey. Now cheap flash (soon the only kind, of course) is short-life. CD's and DVD's fade.

Where do I keep my archive photos now?

3
0

where do you keep data?

In multiple physical locations, on multiple forms of media. Live copy on a mirrored NAS, archive copy on a DVD, another copy using an online 'cloud' (buzzword alert!) backup service and you should be pretty well covered.

3
2
Silver badge

Reading comprehension?

If they are an archive you will presumably write them once and then read them lots of times so you can still keep them on Flash. It is only with many reads that Flash starts to fail.

Or you could keep them in the cloud and let someone else worry about data integrity.

0
1
Coat

@Hayden Clark

> Where do I keep my archive photos now?

How about using silver particles in a thin gelatin layer bonded to paper, glass or a transparent plastic? Ciba and Kodak know about some very stable dyes for colour photographs. Just keep them dry and cool.

Mind you, REWRITABLE CDs and DVDs have very stable dyes too. It's a pity that they don't make them in 12" diameter for bulk storage. (I believe that Phiips trialled modified Laser Discs for data storage in the '80s, but I guess that those would have had stamped foils)

Mine's the one with Kodachrome 25 in the pockets.

2
0

Flash leaks

Flash cells leak, the retention is typically only 10 years, not very long for photos.

0
1

We Are Always Discussing...

...where to keep long/life data. At our house that means photos and video.

Silver images suffer from meltdown when the price of metals gets too high. Ditto rental of Storage either on earth or in a cloud. Urban Legend has it that an NBC exec saved thousands of dollars by throwing out 240 Johnny Carson episodes.

The longivity of writables tends to be related more to the reflective substrate rather than the dies. The use of optical media is about disc size. CD's are a total space waster. DVD's still consume more space than a hard drive. Blue Laser and 12 inch disks sounds very real.

Sony sales people think at least one copy of your TV show should be on ($175 worth of) digital tape. As discussed earlier, a tape dropped out a 6th story window * will probably be more recoverable than a hard drive.

I have not heard much about PROM and ePROM as solid state archival media. If read-while-write, possibly using the image of an electrochemical gap, becomes available, then flash memory would have almost unlimited bit capacity and write cycles per cell, since the gap need only be moved to the next data point, not flushed to one end of the cell and moved back to a data point.

*(no hypothetical pedestrians were harmed in the making of this post)

1
0
Silver badge
Pint

aye

And with a major push to use SSDs in database arrays this could turn out to be a disaster for some people with inoptimised arrays or databases. Especially if they start to cache web front ends on the same array.

Although I see the benefit in terms of power consumption and raw speed, SSDs dont suit all applications.

1
0
Headmaster

The "factor", close to one

Small mistake for sure, but shouldn't the factor not be "the difference", but ... the "factor" or "ratio" ?

"The difference between the amount of data a host wants to write - say 10MB - and the actual data the SSD controller has to write to achieve that - say 27MB - is the write amplification factor."

0
0
Thumb Down

Some minor technical mistakes here...

I know it's not the main part of what you're saying here, but please try to be correct in what you're writing as background material too...

"NAND flash cells have a finite life, in that they only support a specific number of writes": Surely you mean "limited" here rather than "specific"?

"Flash is not byte-addressable, unlike disk drives and DRAM." Last time I checked, disks need to be read/written in (typically 512-byte or more recently 4096-byte) sectors.

2
1
Silver badge

Some real numbers

Here is how to get the number of bytes written per second since the last reboot:

dc -e "512 $(awk '{print $7}' /sys/block/sda/stat) * $(cut /proc/uptime -f 1 -d ' ') / p"

Laptop (web browsing and ssh client): 10521 Bytes per second.

Desktop (media server & software development): 1466 Bytes per second.

The laptop numbers are high because it is either being used or switched off. The desktop is on continuously and also has a magnetic disk for media files (3200 Bytes per second). SSD's might have a limited life on an extremely busy server, but they are fine for laptops and workstations. These two SSD's are too old to support TRIM, and still work fine after 14 months.

Ancient USB flash might have problems because: the capacity is so small that there are not many blocks to spread the writes over, the flash chips might be low quality and the wear levelling algorithm might assume a vfat filesystem.

0
0
Boffin

Erm....

What's wrong with iostat?

0
0
FAIL

i will keep my spinning rust

until SSD actually becomes a good option... do. not. trust.

Fail, because they are going to.

1
0
FAIL

SSDs

SSDs are currently a good option. They have a MTBF that's actually HIGHER than some/most HDDs, and will last far longer than their usefulness, just like my 60GB ATA/100 HDD (that failed btw) is no longer useful compared to a 500GB SATA2 HDD.

Therefore, don't NOT buy SSDs due to "immenent" failure, but rather on their cost/GB vs purpose.

0
0
Anonymous Coward

More pedantry

The phrase "two access cycles are needed to write data to a block that already contains written data" should have read "two access cycles are needed to write data to a PAGE that already contains written data".

0
0
Alert

EVEN more pedantry

In the wear levelling by re-allocation of static blocks that stated a post re-allocation total cumulative writes of 200 is actually going to be 201 as the exchanging of the static blocks will be an additional write for both areas. Just saying:-)

Stuart

0
0
Paris Hilton

Just my two cents...

SSD is just not mature yet. It will eventually become and then I'll get some, but before that time, we are talking the Golden optical disc from Philips in the eighties... They ended up making cd's of which one cannot say they are dead (unless they are (r)w(r)itable) promising but yet not mature enough.

Interesting article though. Always wanted to learn a bit more about SSD's.

Could someone tell this noob what an 80 GB SSD disk should do as holder of let's say my win XP? In terms of lifetime and tools to tell me that the time has come to replace it.

Thank you,

PS: Paris for she knows as much about SSD's as I do.

0
0
Silver badge

Wear levelling is important for SLC too

SLC has an endurance of around 100k cycles meaning that if you keep on pounding on the same physical blocks you can theoretically reach that limit in a few minutes.

SLC just exacerbates the problem because the endurance is significantly reduced.

0
0
Bronze badge
Linux

Me had a flash drive

Died in a few months (9), I'm sure Ubuntu 9.04 contributed a lot to kill it prematurely.

It was pretty fast while it lasted, the computer felt like new.

I think it is time to think out of the box, NAND flash is cool and that, but we need something else more reliable.

0
0
Unhappy

Data security

These wear levelling algorithms for flash drives is why it's very difficult to securely earse data from one. Use something like Heidi Eraser on a file on a flash drive and you can still recover it as the wear levelling has written the zeros / random data to another part of the drive and marked the file's sectors for reuse and later garbage collection. I believe the only way around it is to securely erase the whole drive and not just individual files.

0
0
This topic is closed for new posts.