Posted by bookofjoe 1 day ago
When you create an archive, do not skip the step of reading it back. I'm not talking about just having the system confirm it is readable. I mean actually restoring the data from the tape. I've had tapes fail to retrieve data off of the tape even though the writing app confirmed it was readable.
Drives can have issues reading data written by another drive. Drives can have issues reading a tape it wrote itself.
Why write one when you can write two at twice the time/cost? In all of my experience with tape, I've never written just one tape. In the early days of DVDs, the images were written to LTO to send to have the discs pressed. Two were always sent on the chance one wasn't readable and schedules were so tight that the loss of a couple of days to send a new tape would wreck delivery dates. Another facility wrote three tapes: one kept on site, one sent to IronMountain, one sent to the client directly. Plenty of occasions saw the local copy unreadable and the IronMountain copy was retrieved. These were managed by large robotic tape libraries so each tape was written in a different drive at the same time so it was 3x cost while only 1x time.
Use the drive cleaning cassette as recommended. Your dirty heads can cause the data to not be written/read correctly depending on which head is dirty. If you're having issues reading, clean the heads before giving up on the tape. If you're having issues reading a tape just written, do not think someone else will be able to read it.
Ensure whoever might be using a tape you've written understands what was used to write the tape. Not all archivers are compatible. If you use the tar command to write to tape, people using a GUI app won't be able to read it.
But when you try to read it back with a different, correctly aligned machine, it is unreadable.
In addition the need for cloud providers to comply with userdata erasure means they have to frequently test their ability to read and write their backups - which should increase your confidence in their correctness. Yes problems can and do still happen, but IME far less often than the more traditional tape setups GP describes.
LTO-10 30TB: $259–$300 per cartridge (~$8.63–$10/TB) LTO-10 40TB: $491–$555 per cartridge (~$12.28–$13.88/TB) LTO-9 18TB: $92–$129 per cartridge (~$5.14–$7.17/TB)
https://tapebackup.org/lto-tape-price-trend/lto10-price
One thing worth noting is that the tape cartridges themselves aren't necessarily the biggest expense. Standalone LTO-10 drives are currently listed at around $25,000.
The drives are a fairly fixed CapEx: once they're purchased they don't cost must/anything to run. You may replace them every few years with newer tech for higher capacities of course.
Libraries/robotics last even longer, and can stick around for a decade or more once purchased.
But the fact that -ALL- LTO drive-mechanisms are build by IBM maybe also doesn't help the pricing...
[1] https://docs.aws.amazon.com/snowball/latest/developer-guide/...
The millions of pictures I never got to sort, the multiple versions of a file that I wrote 2 years ago, etc.
But in the view of an AI-company that's probably a paradox. All this data is charged and could be used to learn more about me, it should never be deleted! Instead, some hole should be carved out of a mountain to store it, billing me every month!
Thinking...
User has 15 pictures of his dead mother, deleting 14.
Saved 20 MB! :)
For the pictures, I set up a batch script the one time that used exiftool to read all of my photo's exif data (for creation date), rename the file to that date / time, and put it in a year/month folder structure. Once that's normalized it was pretty straightforward to identify and delete duplicates (assuming the exif data on duplicates has the same date/time in it).
Duplicates are easy, that gets sorted quite fast. Picking 1-2 out of those 5 slightly different pictures of the same occasion is another task.
Shipping your drives to Sam Altman. If everybody did that, he could then deduplicate the pictures of the same thing that multiple people took.
I mean, it's plausible, yet I find it hard to believe this solely drives demand 5 times. Because if it does then what, companies were not using cold storage backups before?
As someone in HPC: people never delete stuff. It can stick around forever, and so if more bits are generated, there are more bits that people don't want to delete "just in case". Depending on the organization, quota enforcement can be tighter or looser.
HSM is fairly popular in the HPC space for this reason:
* https://en.wikipedia.org/wiki/Hierarchical_storage_managemen...
Your newest/hottest bits are on NVMe, anything older than (e.g.) 3 months gets move to SSD, anything older than that to spinning rust HDDs, and then to tape.
There are several file systems where this happens in the background: the file still shows up at /path/for/stuff/foo.txt but the contents are shuffled about behind the scenes.
What I observed in the rising cybersecurity attacks happening around me:
(i) There are organizations thinking that they have [complete] backups and being aware they don't once the cyberattack if effective.
(ii) Even in organizations with complete backups they don't have a [good] process to recover them, and the operations suffered a lot.
Corollary: backups are, in general, not battle tested.
Even within the same company. Every division has backup processes, only few actually try to recover the data once in a while...
I've ran drills with companies which concluded that the incremental backups couldn't be restored due to some data corruption issue (no readback test was ever done).
Luckily the company also did a full backup every 6 months, so a disaster would have only thrown them back 6 months at max! /s ;)
But yes, some issues were even more simple: I had a case where performing a restore required twice the storage size of the uncompressed data, and the company actually didn't have the infrastructure ready to do that in case of a disaster...
Don't get me wrong, drills are important! You should do them. However backups without a restore drill better than no backups at all. Even if you only get the last full backup, the data from the last 6 months will be the easiest to recreate from scratch if you must.
I'm suspecting the AI companies simply save everything (even if you said "no"), just in case they need it later for training.
Why? Because there is no moral anymore. They blatantly ignored copyright after all.
As time goes by, people are sending more and more. It started with just prompts, then the local source code, then calendar, emails, local files...
At some point in the future the amount of data being stored indiscriminately must consume such a huge amount of resources and cost, that it becomes a hidden fixed cost within a country's GDP, without any tangible benefit...
That continues to be the assumption, yes. Disregarding all the waste being created in the process.
What's fascinating is the fact that the production of TAPE sees a 5x boost. That's the format with the slowest retrieval rate, only making sense for long-term storage, usually 1+ years.
So is there really 5x more valuable data now, or is it just not economic to actually judge the need for preserving data...?