It’s RAID. With Floppy Drives.

There are some tings that should be possible, so just have to be tried. [Action Retro] has a great video showing just such an escapade, the creation of a large RAID 0 array using a pile of USB floppy drives. Yes, taking one of the smallest and most unreliable pieces of data storage media and combining a load of them together such that all the data is lost if just one of them fails.

Surprisingly the process of creation is quicker and simplier than we expected, with a slightly long-in-the-tooth version of Mac OS X making short work of the process. Starting with 30 USB floppies and a pair of large USB hubs, he whittled the pile down to 13 drives that would play nicely and RAID together. The sight of so many drives all lighting up together as the precious megabytes are filled with data is probably not one seen outside the realm of floppy duplication machines, which brings back bad memories for those of us in the consumer software business in years past.

Would you do this? Probably, but should you do it? Of course not, but then again he’s done it so the rest of us don’t have to. Here in 2022 maybe there are better uses for a brace of floppy drives.

38 thoughts on “It’s RAID. With Floppy Drives.

  1. Sometime in the grey dreariness of a 1997 northern winter, I managed to configure a pair of floppies into a single volume that delivered data at twice the speed. This wasn’t the objective, I think I was trying to figure out a 4 drive configuration by jumper setting and i/o address manipulation. There was also drvparm involved I think, due to somewhat concurrently investigating extra tracks and weird disk formats. i.e. I might also have had some drive parameter settings loading that made this happen as well as odd hardware config. This was under DOS 6.2. I do remember I also put a compressed volume on top of the 2.8MB volume it made and got about 3.5MB out of it.. and it was faster again. (Drive compression is always faster when your CPU is idle while waiting for relatively slower storage. fast storage, slow CPU don’t bother.)

    I wish I had notes on this so I could revisit it. I had a number of strange software and hardware factors in play. For instance I was also screwing around with a util that made an Amiga volume file on a PC hard drive, which I found one could also mount and format to FAT16 and have a movable virtual volume…. annnd I might have glommed that on top of it to make a contiguous volume in the first place, IDK.

    1. > at twice the speed

      that could only happen with 2 independent floppy controllers and custom floppy driver/low level access.
      -µPD765A controller has only 1 input, meaning only 1 floppy can be active at a time
      -DOS uses DMA 2 and IRQ 6 and those cant be shared between two controllers working simultaneously

      I only read about one program capable of that – SYDUPE, custom and highly specialized disk copier targeting disk duplicator/pirate market?

      1. Well I’ll correct you there, it only knows/intends one floppy is active at a time. I definitely had them on the same motor active line or something.

        I was incredulous about it actually reading them independantly, that’s why I stuck a compressed volume file on it, to see if it was actually twice the space and not just 1.44 MB confused between the two of them. CVFs find out about any disk flakiness quite quickly.

        At the time though, it was tying up my floppy drives and floppy interface, no network installed, so there wasn’t gonna be any useful utils to check things out going onto or off the system. It was one of those, play with it on the weekend, but Sunday afternoon it needs put back to rights so you can do work. I think I thought I was going to remember what I did, and do it again. But it was one of those things where you go seeking a guru and just get a bunch of knowitalls. I have a suspicion I might have bugged Alan Cox about it at some point. Anyway, it stretched out such that I didn’t get any enlightenment, I got busy on other stuff and forgot what the hell I did, and even to retry it until years later.

        I think they were both JU-257s, older ones with a fullish jumper block drive IDs 0-3 and 4 to 6 other configuration jumpers.

    1. CD-R wasn’t too bad, provided you bought decent quality discs (medical grade were good), and kept them out of sunlight. And in the early days didn’t burn them too fast. In a previous job we used hundreds every week so we learnt a bit about them.

      CDRW was useless. Write once, read never. We tried it to save costs and waste, but it was useless.

      1. Burn slow, store out of light, not in a cardboard sleeve (acidic paper) nor against vinyl plastic, never write anything on them even with CD markers, keep out of humidity, not too cold, not too hot. Do all that and half of your discs might survive after 5-10 years of storage.

        The earlier 2X discs were fine, but the later drives were actually too fast to burn them properly.

        1. I don’t know what is wrong with your CD drives. I have CD-Rs I burned in 2004 and they still read back just fine. I have burned CD-Rs from the 90s that still work fine too. I used to keep them in jewel cases or sleeves, but these days they’re all in spindle cases.

          My floppies, on the other hand, have not reliably held for that long.

          1. >My floppies, on the other hand, have not reliably held for that long.

            Some time ago I found my old floppy keeper with some diskettes dating back to 1992, including some old photo floppies from when they were trying to introduce the concept. 50 box of diskettes, maybe 98% of them worked flawlessly and the rest were recoverable. Not much interesting on them though.

            When I was a kid, we’d have boxes of floppies stuck in our pockets to copy games off of friends. They were not well cared for, cheap off-label disks, and with a compressed archive spanning 10-20 diskettes you’d often get one diskette that gave you read errors – but you’d always get the file out somehow and then scan/reformat the diskette and put it back in the box for use. Mostly the problems were with the drive not aligning properly, not the diskette.

        2. I still own today (2022) my first 1997 CD-Rs burned with my first parallel external 2x-2x6x burner (2x R write -2x RW write-6x read) and PC PIII-333 Win 95 – Nero Burning and I have them read without problems in 2021.

      2. >(medical grade were good)

        The problem was that it all became a Lemon Market (look it up). Manufacturers would claim 100 year data retention, with no way for consumers or even professionals to verify the fact (obviously), and the only warranty on offer would be that you get another disc. Good luck trying to prove that the disc was defective.

        With no way to differentiate between good discs and bad discs before purchase, by brand or otherwise, consumer trust went down and people would not pay a cent extra for proper discs, so none of the brands even bothered to make good discs. Consumers demanded fast and cheap throw-away coasters, so that’s what they all did.

        The dyes used on CD-Rs are inherently unstable with a limited shelf life. Counterfeit discs were also common. Another problem was that they could do accelerated aging tests on randomly selected discs out of a batch, but could not reasonably guarantee a lack of defects even for the expensive discs because all possible tests you could do on a disc were destructive, so they were simply not suitable for archiving.

        1. The “tail emissions” of the floppy media market were much like that also. Got two lots of Memorex disks, pre millennium and post, and they are chalk and cheese.

          1. Yeah, though magnetic media doesn’t have a “best by” date in the same sense, and you can test a diskette for recorded signal quality non-destructively so you can actually go through a pack of diskettes and throw away the dubious ones.

          2. Another problem was that when you burned a disc, it would come out with data errors from the start due to dust and vibrations and defects on the disc. The error correction codes helped to mask these errors, and burning slower would reduce the error rate, but practically speaking you could never burn a perfect CD-R. The maximum allowable block error rate for a CD-R was specified at 220 per second and a “good” disc would come out below 50/s.

          3. One important difference between a CD-R and a diskette is that the diskette was master formatted in the factory. Nobody sold just plain blank diskettes because the expectation was that you’d just pop one in and start copying files.

            The CD-R had a pressed tracking groove, but it was not tested to read back what was written on it for obvious reasons, while the diskettes were already use-tested out of the box and disks which failed to read back properly were mostly eliminated at the factory.

      3. I remember when CD-R could become toaster real easy in the early time, before anti-skip or anti-burn technology was improved. Anything like inconveniently timed virus update notification, a stutter in Solitaire, or even just bumping the computer case would cause coaster. And back then, they were averaging $1 a disc new in bulk so it was sort of expensive with US minimum wage of $4.25 per hour. (after paycheck taxes plus taxes on new CD-Rs, you could only afford 2 coasters per hour!)

        DVD weren’t without problem, there were matter of DVD+R and DVD-R, the drives weren’t interchangeable with those formats initially so you had to choose the right one. And I remember the first time DVD-RAM came out, $25 for one yikes! Those wouldn’t work in vanilla DVD-ROM at all. I pretty much skipped burnable BD in favor of cheaper and more reliable USB storage and I haven’t had any working optical drive for any of my computer in 3 or 4 years now.

        Today I can get memory stick that doesn’t becomes toaster, can work with just about any computers made in the last 25 years, and can be reused a lot without any issue.

        1. Solid state media too has limited data retention because of cosmic radiation, read disturbance, and flash memory being inherently “leaky” – the gate charges dissipate over time, especially at elevated temperatures. I would not trust a random USB stick to keep all of its data for 10 years.

    1. Commodore disk drives could have 4 drives total via hardware ID selection and if you ran software ID selection to disk drive, theoretically you could have 247 total disk drive. Realistically, serial communication can get unreliable over long distance since there was no signal buffering or amplifying. I was able to get 8 disk drives working at once in early 2000 after Goodwill had loads of used C64 stuff on sale cheap. The printer wouldn’t work though due to obscenely long serial connection. It may be possible to do better with easy to find 6 pin DIN plugs and 6 wires shielded cables, make really short custom cables to reduce signal issue and cram more. I think I have total of 15 working 1541’s and 2 of 1571’s plus one more built in my C128 but power would be another issue, 25w each. Lights would dim if I ran all those disk drives at once with their stock and inefficient power supply.

      Realistically most games were designed for 1 or 2 disk drives. It would have been nice if Ultima 5 did support total of 8 disk drives since they used 4 disks, both sides. No more swapping at all. Ultima 5 supported 2 drives max and I have no idea if anyone ever hacked it to support more.

  2. I was a student at UC Berkeley in the days of the Mammoth project which was part of the early development of RAID storage techniques. Used to have labs in Evans hall and I would walk by an informational display where a research group showed how they were testing early RAID-like setups with big stacks of floppy drives just because they were cheap and accelerated the failure timeframe. At that time they were still 5¼” drives if I remember correctly.

  3. IBM had a storage system mate up of rolls of tape (looked like toilet paper) and a motion system that would store and retrieve the tapes from a giant storage unit. Great for large data, not great for speed. Remember a tractor trailer full of CD-ROMS going 60 miles per hour is many, many gigabits per second.

    1. Oh dear LORD the Mass Storage Subsystem (MSS)!!! It was two tape cartridges=one 3350 disk drive. It was part of the Heierarchical Storage Manager architecture. You’d “mount” a virtual disk drive and the MSS would load its image off tape to a staging disk, then write it back to the tapes after you dismounted the virtual disk. We had one of those where I worked back in the early 80s. It had a painfully high failure rate (tapes would get snarled up in the drives, stuck to the capstans or the head drums) until IBM sent a small platoon of engineers to our site and they finally determined that it was airborne contaminants (mostly very finely atomized oil from air handler motor bearings or somesuch). They raised the temp in the machine room and/or started using higher grade filters in the air handlers and the problems largely went away. It was eventually outclassed by robotic tape libraries that could handle the up-and-coming 3480/3490 tape cartridges.

  4. I’m going to go out on a limb and assume the first sentence was meant to be this:

    “There are some things that seem to be so impossible, they just have to be tried.”

    I had to read it 4 times and still didn’t understand. Not being a grammar NAZI or anything. It was just very confusing at first. I know English is not the first language of many so I’ll apologize in advance if anyone feels insulted. It isn’t intended to be insulting.

    1. I think the first sentence works, but I parsed it as:

      “This project has never been done before but it technically *should* work, so someone *had* to try it to prove that it could be done.”

  5. I played with something like this in the early naughties, but for a different use case; Data resiliency on sets of floppies.

    I got it to work well enough doing RAID-5 on n+1 disks, even played around with doing it in n (where n>9) disks by playing with disk geometry and writing 1,600k to the disks.

    Never got to use it though, the company I was with decided that they’d just start shipping data updates on mini-CD.

    My next toy, the CDROM RAID-0, actually got put into commercial use by an advertising company. They had to send gigabytes of data between the coasts, were leery of sending client data over the internet, and FedEx kept breaking tapes and hard drives.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.