Hard drives should not be left unattended for even a few years with any expectation that they will still function when plugged back in. Sure, many will, but it's a crap shoot: Under ideal conditions an unused hard drive can be expected to retain data for 9 to 20 years. [0]

[0] https://blog.macsales.com/43702-we-bet-you-didnt-know-that-y...

I really wish this blog would have listed some sources for all these statements.

Well yes but I heard that b4. You should just put on your calendar to plug them in (and update the backup) once in a while. I do not think it is about mechanical robustness. It is about the way the data is stored with magnetic tech, it think HDDs should be operated every once in a while to have a better chance of not loosing the data. But 9 days sounds awfully short.

I found the approximate time frame in a few places, but never with any primary source. I also came across the claim that hard drives used to be more mechanically robust, and perhaps failed less often for such reasons. Again, I couldn't source it though.

Well, yeah, the idea of individual drives is great, but in practice I've had a lot of drives die.

This was mostly a cost comparison I wanted to show, as the solutions of rolling your own set up at home vs a cloud set up are definitely very different.

At 10TB, you're paying $600/year on B2, a price at which you could outright buy two 10TB drives and stick them in a larger NAS setup with ZFS or a simpler RAID-1 array.

That's reasonable. Alternatively, I'm using Arq[0] (locally encrypted before upload, uploads to tons of different services, I'm using AWS S3 and Google Storage) so I'm paying ~$7/month for about 40G stored. And I don't have to worry about local drives dying or being stolen. Just another option.

[0] https://www.arqbackup.com

> locally encrypted before upload, uploads to tons of different services

https://www.duplicati.com does that for free being open-source thing, all platforms supported.

Don't forget B2 involves putting that data onto multiple hard-drives and ensuring it isn't lost. What if that single disk were to be lost? You're paying Backblaze to ensure that GB can be accessed in the future

> What if that single disk were to be lost?

This was a cost analysis, mostly for those that can't afford B2. You can buy two drives for the cost of a year at Backblaze and have N+1 redundancy.

> You're paying Backblaze to ensure that GB can be accessed in the future

No doubt, which is why I prefaced this with "If you don't need to have your data in the cloud".

Having tried to keep "permanent" backups of my personal digital data for 35+ years, I should point out that one should not underestimate the problem of technological change.

The "device" you use today will, in all probability, not be usable after a number of years, no matter whether you spin it up regularly and maintain it in a working condition.

I have backups in countless floppies, diskettes, Zip drives, 9-track, QIC, DAT, SCSI, ...

B2 costs more than that, unless you never download anything. (Not taking into account exploits like utilizing Bandwidth Alliance and Cloudfare free tier to get free egress.)

How is that exploiting anything, they encourage it.

It's against the Cloudflare ToS:

> The Service is offered primarily as a platform to cache and serve web pages and websites. Unless explicitly included as a part of a Paid Service purchased by you, you agree to use the Service solely for the purpose of serving web pages as viewed through a web browser or other functionally equivalent applications and rendering Hypertext Markup Language (HTML) or other functional equivalents. Use of the Service for serving video (unless purchased separately as a Paid Service) or a disproportionate percentage of pictures, audio files, or other non-HTML content, is prohibited.

It’s okay at least for serving images (as long as you use workers): https://news.ycombinator.com/item?id=20791660 from CloudFlare CEO. It’s not clear to me if piping terabytes over terabytes of, say, opaque data would violate anything.

This is a clever exploit none the less (not a pejorative).

> then putting them ["a regular HDD"] in a box, no need to keep them spinning

Are you sure about that? I'm no expert, but I thought it was advisable to power up spinning platter disks at least every so often, once a year say?

And where are you keeping this backup, same box?

> but I thought it was advisable to power up spinning platter disks at least every so often, once a year say?

Is there a specific reason why you would need to do that? I'm not really aware of any.

> And where are you keeping this backup, same box?

This is ultimately up to you, depending on what kind of redundancy you deem necessary and whether you have multiple physical locations where you could store the said drives.

These days, hard drives probably scrub themselves periodically, looking for and rewriting sectors with high error correction error rates.

SSD’s definitely do this.

Even if hard disks don’t do this, mean time to repair is a crucial parameter in raid durability calculations. If you spin the drives up once a year, the mean time to repair is 6 months. Most raid systems are likely to lose data if it takes more than a few days to repair.

The ball bearings have lubricant that can shift over time due to gravity. And the metal parts might weld to each other over time if they are not moved.

How likely are these things to actually happen?

I think the welding takes more than a year. The lubricant settling probably depends on the quality of the drive and the ambient temperature. And I guess the ball bearings would be pretty tolerant (since you only have to move them a little to get the lubricant on them again), it's more likely that the heads would stick.

https://en.wikipedia.org/wiki/Stiction#Hard_disk_drives