Hacker News Reader

Comment by shiroiushi | Hacker News Reader

There was a period of time where pressed CDs were manufactured poorly, with the aluminum layer inside exposed to the outside, resulting in corrosion over time and loss of readability on those CDs.

Overall, though, properly-made CDs, handled carefully, have been excellent at storing data long-term.

But while this is nice enough I guess for storing individual musical albums long-term, it's not practical for storing truly large volumes of arbitrary data. CD-Rs and CD-RWs have not had the same durability demonstrated at all (quite the opposite in fact). DVDs are better at almost 4GB per disc, but here again only the factory ones are actually durable, and 4GB isn't going to store much these days, perhaps one movie with high lossy compression.

View on HN

While my comment wasn't about the feasibility of pressed CDs for a mass blackout event but just an example of long-term integrity of existing digital media, it's unfortunate that a forum (MyCE) dedicated to tracking integrity of user-writable optical discs unexpectedly closed a couple years ago due to the webmaster pulling the plug.

It had users who carefully performed benchmarks on media for more than a decade to see which types and makes held up best over time, along with best practices. Few have the interest or patience for such things so it's unfortunate to just have such info vanish.

I will add though that what's missing from the discussion is Blu-Ray, which allows up to 128GB per disc. (I only vaguely recall reading some critique of BD DL discs so can't say how it might compare long term though, apart from the greater cost at such capacities.)

Somehow it seems ironic that a forum dedicated to understanding the long-term viability of data storage, an important topic lately because of the unreliability of 3rd-party providers (like cloud companies), itself became a victim of the unreliability of its own webmaster.

128GB BD-R discs do exist, but at $219 on Amazon for 25 discs, that's about $0.07/GB. It would be MUCH cheaper to just buy a stack of refurbished enterprise-class HDDs and store your data on those, in triplicate, with a filesystem that has error correction (like ZFS). Personally, I would bet on HDDs used this way still being readable and not having bit-rot after 50 years over 4-layer BD-R discs.

Wouldn't it be more like $0.07/GB for BDXL? If one got particularly lucky with HDD failure rates perhaps they'd survive running that long in RAID but one would expect some replacements over such a long period.

Some other things to consider are at high capacities all HDDs use helium now, which slowly leaks (WD/HGST have a SMART stat about the level*) and the cost of running drives/associated computers/maintenance over a long span vs the up front cost of passive writable media (edit: for some reason I assumed this was what was meant but they could be left cold which would likely increase survival odds and be cheaper).

* And there isn't much long term data about it that I could find, though some have reported between 1-5 years the SMART stat either remaining at max or dropping a few digits. Even Backblaze outside of their first article a year into using them hasn't seemingly continued reporting on the stat that I've noticed. I get the sense though that other types of failures are expected sooner than leaks.

>Wouldn't it be more like $0.07/GB for BDXL?

Whoops, thanks for pointing out the math error; I've fixed it.

>If one got particularly lucky with HDD failure rates perhaps they'd survive running that long in RAID but one would expect some replacements over such a long period.

I don't think so: I'm not talking about keeping these drives spinning for 50 years, but rather in cold storage, just as we'd do with the BDXL discs.

>Some other things to consider are at high capacities all HDDs use helium now, which slowly leaks (WD/HGST have a SMART stat about the level*) and the cost of running drives/associated computers/maintenance over a long span vs the up front cost of passive writable media.

Helium leakage is an issue I didn't think of, and I don't know how sitting in cold storage for 50 years would affect this. But again, the costs of running drives/maintenance/etc. should be zero, because I'm comparing apples to apples. No one would seriously propose a massive array of BDXL drives with BDXL discs continuously available, so likewise I'm proposing just keeping 10+TB HDDs in cold storage.

Reply on news.ycombinator.com