Seeing these errors means “the SSD is on its way out,” according to HTWingNut.
Since we’re simply talking about being unpowered for a while, wouldn’t a simple full format fix/reset all ECC errors? No need to scrap the drive.
Surely a cap/transistor temporarily losing charge shouldn’t permanently destroy it!
Anyways, HDD for 6-24 months offline data storage, SSD for always-online data storage, and flash if you’re a masochist like me.
I think tape storage has the best longevity in offline data storage, but it’s been a while since I checked.
Yeah I believe tape is still king there. LTO is working on some 500+ TB tape for the future IIRC.
The upfront cost of tape is excessive though. It wasn’t always like that. And LTO-9 missed its capacity target: it’s 18TB (1.5x LTO-8) instead of 24TB as planned. Who knows what will happen later in the roadmap.
They’ve missed a couple of times over the years.
From LTO 1 to 9, the capacities (TB) were 0.1, 0.2, 0.4, 0.8, 1.5, 2.5, 6, 12, 18. LTO 6 also rather let the side down there.
Apparently though LTO 10 is going to get things back on track? I’ve seen claims it will achieve 36TB, but I’ll believe it when I see it.
The real problem is the environmental requirements for LTO 9 and newer have become too strict. The longevity is still (supposedly) fine, but the tapes are much more sensitive to temperature and humidity fluctuations when in use.
Brand new tapes have to be brought into the environment where they’ll be written for 36-48 hours to acclimatise before being used, and then have a 60-90 minute “calibration” in the drive before they can be written to.
Honestly, it could put the use of the newer types of tapes entirely out of the reach of many.
Oh I didn’t know about the new requirements. Less backwards compatibility too. IBM 3592 looks better but costs even more. Tape drives can’t be that much higher tech than HDDs, so if they cranked up the volume they could likely be way more affordable.
Fascinating!
Strictly speaking, I think paper beats magnetic tape on longevity.
Unfortunately, it loses on data density.
If we are going by that metric clay tablets beat paper.
I was excluding media that are impractical for most people to use.
Paper would fall under that these days, wouldn’t it? You can’t just fit a word (8 bytes) onto a punch card like the old days, and you’d need billions of the things go even start matching up to modern storage.
I did call out data density in my first comment. Did you somehow miss that? Not all things that need storing are megabytes in size, though.
Why would you assume that paper means punch cards? Printers can store far more than a machine word on a page, are relatively cheap, and are widely available. For some things, this can be superior to both magnetic and flash storage.
Not to mention PaperBack, it’s essentially making a backup for life.
Depends on the threat model and how long do you need the data.
Worked on a place long ago, that anything they needed to save offline from more than a few decades where stored in microfilm, the expectancy there where they would last 80 to 100 years.
Anything else was pretty much tape.
You also take in account the technology avaiability. The more complex is to use, harder will it be to reproduce in the future. Even with tapes, you might want to copy the data to another tape/recorder every decade or two, to keep it on par with the technology.
I etch my data only metal slabs. The longevity is great, but the bits per pound is rough.
Just make sure you dont use sub-standard copper. Future generations thank you
iunderstoodthisreference.jpg
At least you didn’t have to worry about Ruin manipulating it.
… Sorry, I’ve been reading the Mistborn series and just finished “The Hero of Ages”
instructions unclear. I now have railroad spikes in my eyes for some reason.
Technically does, but has a very high rate of failure on recovery, you need to recover the entire drive not just a section, and it can take days or weeks to read it back, vs mere hours.
Tape presents its own share of problems. If not strored in some very particular conditions, like temp, humidity, and others that I can’t recall, they can stick to tbe adjacent layers, become brittle, curved, etc…
M-Disc
As we all slowly step back from CD-RW, no quick movements… or was that all just a bad dream?
https://en.wikipedia.org/wiki/M-DISC
M-DISC’s design is intended to provide archival media longevity.[3][4] M-Disc claims that properly stored M-DISC DVD recordings will last up to 1000 years.[5] The M-DISC DVD looks like a standard disc, except it is almost transparent with later DVD and BD-R M-Disks having standard and inkjet printable labels.
In 2022, the NIST Interagency Report NIST IR 8387[25] listed the M-Disc as an acceptable archival format rated for 100+ years, citing the aforementioned 2009 and 2012 tests by the US Department of Defense and French National Laboratory of Metrology and Testing as sources.
That being said, that’s 100GB a disc. You can stuff a lot more on a typical hard drive, and I appreciate that people want to easily and inexpensively reliably store very large amounts of data for the long term.
EDIT: At least in a quick search on Amazon, while there are plenty of drives rated for M-DISC, I don’t see any kind of “take hundreds of discs, feed them mechanically in and out of a drive” device that’d let one archive very large amounts of data automatically. You’d need 100 of those to fully archive a 10TB hard drive.
I’m sure USB pen drives are even worse.
I actually just pulled some files off of one from 2004-ish. No issues. Found another one from 2008 about a year ago that had no issues as well. Not sure why… maybe because they were so much lower capacity? Like, one was 64MB and that was huge back then.
They were slc, so the charge ratio was much higher.
Mlc/tlc/qlc drives have to measure a current very precisely, up to 16 values of discrimination, any charge degredation doesn’t change a 1 to a 0, but a 3 to a 2 to a 1 and given enough time, a zero.
Also smaller gate dielectric so more leakage.
Those old drives may be using SLC flash. It can have a 20+ year data retention.
I pulled some data off some old Samsung 1TB SSDs that werent powered for 3-4 years without an issue either. I guess they were SLC based on what others are saying.
I guess it’s a your mileage may vary situation depending on the exact drive you purchase and probably other factors too.
The four tested ‘Leven JS-600’ branded SSDs are basically bog-standard no-name units.
Oh.
Bog standard? They are more like “god knows what’s inside these” and cionsidering I have a solid suspicion there is no god…
There is a fairly reasonable theory floating around that no name drives have B quality chips, so these may have started with chips that were iffy from the start. Id like to see a test of this type , carried out by Backblaze, with thousands of drives.
Reject flash, return to tape
The tape is a lie.
I’m making a note here
Does nvme have this issue, too?
Absolutely. Sata SSD, m.2/nvme, USB thumb drives, it’s all just different form factors for nand flash memory.
Figured as much, but I wasn’t sure if thre nvme flash was of higher quality with potential benefits like what SLC brings. Thank you.
Heh, well some of those are most definitely of higher quality, but you mostly see that difference in throughput and seek times. But the underlying storage mechanism is the same, so yeah, this aspect is probably universal.
