[EvanAnderson]

That's sad, but it mirrors my experience with commercial customers. Tape is so fiddly but the cost efficiency for large amounts of data and at-rest stability is so good. Tape is caught in a spiral of decreasing market share so industry has no incentive to optimize it.

Edit: Then again, I recently heard a podcast that talked about the relatively good at-rest stability of SATA hard disk drives stored outdoors. >smile<

[duskwuff]

Tape is also an extraordinarily poor option for a service like Internet Archive which intends to provide interactive, on-demand access to its holdings.

[BryantD]

Back in the day, if you loaded a page from the web archive that wasn’t in cache, it’d tell you to come back in a couple of minutes. If it was in cache, it was reasonably speedy.

Cache in this case was the hard drives. If I recall correctly, we were using SAM-FS, which worked fairly well for the purpose even though it was slow as dirt —- we could effectively mount the tape drive on Solaris servers, and access the file system transparently.

Things have gotten better. I’m not sure if there were better affordable options in the late 1990s, though. I went from Alexa/IA to AltaVista, which solved the problem of storing web crawl data by being owned by DEC and installing dozens of refrigerator sized Alpha servers. Not an option open to Alexa/IA.

[stonogo]

This is a common use for tape, which can via tools like HPSS have a couple petabytes of disk in front of it, and present the whole archive in a single POSIX filesystem namespace, handling data migration transparently and making sure hot data is kept on low-latency storage.

[BryantD]

Yeah, it was like this (except not petabytes).

[EvanAnderson]

I presume backing-up the archive is a desirable thing. That's a place where I would see tape fitting well for them.

[duskwuff]

Perhaps? But unless tape, and the infrastructure to support it, is dramatically cheaper than disk, they might still be better served by more disk - having two or more copies of data on disk means that both of them can service load, whereas a tape backup is only passively useful as a backup.

[stonogo]

    unless tape, and the infrastructure to support it, is dramatically cheaper than disk, 

This turns out to be the case, with the cost difference growing as the archive size scales. Once you hit petascale, it's not even close. However, most large-scale tape deployments also have disk involved, so it's usually not one or the other.

[xk3]

You might squirm at using refurbished or used media but those 3TB SAS ex-enterprise disks are often the same price or cheaper than tapes themselves (excluding tape drive costs!). Will magnetic storage last 30 years? Probably not but they don't instantly demagnetize either. Both tape and offline magnetic platters benefit from ideal storage conditions.

[EvanAnderson]

It's not just cost / media, though. Automated handling is a big advantage, too. At the scale where tape makes sense (north of 400TB in retention) I think the inconvenience of handling disks with similar aggregate capacity would be significant.

I guess slotting disks into a storage shelf is similar to loading a tape changer robot. I can't imagine the backplane slots on a disk array being rated at a significant lifetime number of insertions / removals.

[stonogo]

If you're ok with individual storage units as small as 3TB, then we're talking about a different set of needs. At that scale, whatever you can lay hands on is probably fine. Used tape is also cheaper than new. IA is dealing with petascale, which is why I mentioned that the price difference widens with scale.

[EvanAnderson]

A lot of people, me included, consider anything online not to be backup. Being disconnected and completely at-rest is a very desirable property.

[Melatonic]

That's exactly what it is.

You also don't want your true backups online at all - that's the whole point.

[Melatonic]

Tape is almost always used for cold storage backups that are offline in case of ransomware attacks. Using it for on demand access would be insanely slow