This is, quite possibly, one of the best nerd sniping comments I’ve seen.
Two thoughts come to mind, one not serious, one serious.
1. I can’t imagine having to align the counters on all those tapes.
2. I’m guessing this would really only work for sequential reading and writing. In some ways that makes it more fun as the latency would be that much worse.
I have done it with usb floppy drives under openbsd, I am sure it is just as trivial under linux but I had obsd and a bunch of usb floppy drives at my disposal.
#it has been a few years I don't remember if it works with bare drives or if you need a disklabel on each floppy
bioctl -c 5 -l /dev/sd2c,/dev/sd3c,/dev/sd4c softraid0
#the raid will show up now, check dmesg
disklabel -E sd5
newfs /dev/sd5a
mount /dev/sd5a /mnt/floppy/
umount /mnt/floppy
bioctl -d sd5
#after inserting all floppies reassemble the raid
bioctl -c 5 -l /dev/sd2c,/dev/sd3c,/dev/sd4c softraid0
mount /dev/sd5a /mnt/floppy
I love it when a system like this A. does not try to railroad you into the "correct path" and B. the independent layers actually work independently.
One day in what was probably sullen resentment that openbsd has no equivalent to DRBD I assembled a raid on iscsi drives, that is, initiate 5 iscsi sessions to independent hosts then assemble a raid with them. and you can imagine my surprise when it very nearly worked, I could read and write just fine. The part that did not work was drive failure. My guess is that iscsid did not fail a drive in a way that softraid understood. so a drive failure just lead to everything hanging.
> I love it when a system like this A. does not try to railroad you into the "correct path" and B. the independent layers actually work independently.
Agree! My first thought while reading the article was that it would be very easy to do this on OpenBSD as well, either with USB floppies or normal 34-pin drives as well.
OpenBSD's softraid stack doesn't care much about what the underlying hardware is as long as it looks like a disk and talks like a disk.
There is a YouTube video on the Action Retro channel, where this article is used as inspiration. Apparently you're not able to use any random floppy drive, but you can use more than five.
I thought about trying this with LTO drives, to have a ridiculously slow but also ridiculously high capacity raid, but sadly the LTO tape decks are a bit too expensive for this experiment.
Not floppies but I clearly remember some Sun Microsystems video demonstrating ZFS where some guys dressed as over the top engineers randomly disconnecting USB thumb drives that were part of a pool to show the file system resilience.
This feels like a storage solution that needs a "|0| days since last data loss" sign. Take the reliability of floppies under continuous read write cycles and divide it by 5?
Readit News
Action Retro has a video with floppies: https://youtu.be/1hc52_PWeU8
He also references a MacWorld article with Daniel's array: https://www.macworld.com/article/165663/floppyraid.html
Need to see if someone can hack together RAID on cassettes on an old Apple II
Two thoughts come to mind, one not serious, one serious.
1. I can’t imagine having to align the counters on all those tapes.
2. I’m guessing this would really only work for sequential reading and writing. In some ways that makes it more fun as the latency would be that much worse.
One day in what was probably sullen resentment that openbsd has no equivalent to DRBD I assembled a raid on iscsi drives, that is, initiate 5 iscsi sessions to independent hosts then assemble a raid with them. and you can imagine my surprise when it very nearly worked, I could read and write just fine. The part that did not work was drive failure. My guess is that iscsid did not fail a drive in a way that softraid understood. so a drive failure just lead to everything hanging.
Yeah:)
> Unix was not designed to stop you from doing stupid things, because that would also stop you from doing clever things.
- Doug Gwyn
Agree! My first thought while reading the article was that it would be very easy to do this on OpenBSD as well, either with USB floppies or normal 34-pin drives as well.
OpenBSD's softraid stack doesn't care much about what the underlying hardware is as long as it looks like a disk and talks like a disk.
https://www.youtube.com/watch?v=1hc52_PWeU8
Now if you are stripping... Well... then sure the data loss is your own fault, you have taken the R out of RAID.