The scale usually tips towards LTO at around ~100TB or so (give or take 20TB depending on where you are located and how good your negotiation skills are), and that's only if you can afford a tape downtime until you get a new one (in case it breaks), and to refresh the whole library come LTO-10 (because it won't even be able to read your LTO-7 tapes).

It almost always makes better sense to back up to hard drives that carry their own connectors and "tape" with them -- I've recently found two disks from 2001 in storage. I needed a $15 IDE-to-USB connector because I don't have IDE connectors anymore. And there were 16 bad 512-byte sectors over one 20GB disk, and no bad sectors over one 15GB disk. Easily readable.

I also had a couple of backup tapes from the same era inside the same box I found. The required drive brand is not even listed (it's not LTO), and I'd be surprised if (a) I can find a tape that can read them, and (b) if they still work properly.

But I don't need them, because the drives they backed up are still readable .....

Tapes are only good for offline storage.

If you want online storage, you'll need one drive per tape, and you would get huge lookup times.

> Tapes are only good for offline storage.

> If you want online storage, you'll need one drive per tape,

No, tape libraries exist, and would be considered 'nearline' storage.

https://www.ibm.com/it-infrastructure/storage/tape/library

> and you would get huge lookup times

Yes. This is not a good system for random access.

Indeed. But the initial investment is very expensive.

It depends on your financial scale. A fancy tape library, sure, but if your company is in-sourcing hundreds of TBs of backups surely you are not destitute. Glacier is always cheaper.

On the other hand I paid $185 for a LTO-5 drive a year or two ago (no library, manually operated). 1.5 TB on a $20 tape is still pretty good in my book.