MFMulation? (Solid-state replacements for MFM drives)

On 4/16/07, Alexandre Souza <alexandre-listas at e-secure.com.br> wrote: I am aware that CF does "wear levelling", but there is a substantial difference of quantity of write activity between, say, CP/M, and VAX/VMS. With CP/M, if I'm sitting at a prompt or sitting inside Wordstar, I don't think the OS is writing to the hard disk behind my back. With VMS, it's pretty much constantly writing to the boot volume at least (you could mount write-protected data volumes) whether you are typing any commands or not. In the case of CP/M, you could decide to accept a 1GB card and present it to the host as, say, a 5MB drive. That would be rather a lot of space to a CP/M machine. You'd be able to fill that drive 200 times, in theory, before you reused a single segment of the CF card. In the case of VMS, though, a 5MB boot volume isn't particularly useful. The smallest I've ever shoehorned VMS 6.0 into (and it was a trick) was an RD54 at 154MB. Practically speaking, unless you are going to use it only with pre-VMS-5.0 distros, you are going to want to present a drive to the VAX in the hundreds of megabytes. If you decide on, say, 250MB, that's quadruple sector redundancy for complete drive re-writes, but in practice, only a percentage of the drive is going to change on a frequent basis as the machine runs. Since the OP mentioned VAX/VMS and I'm reasonably certain that the target would be a MicroVAX with either an RQDX1, RQDX2 or RQDX3, or the near equivalent in an MicroVAX 2000, it's likely that the RQDX controller is the only drive controller on the system. Since part of the problem with machines of that vintage is the lack of available, reliable, large ST506/ST412-interface MFM drives, the whole point of drive emulation is to take a MicroVAX and get it running with no real MFM drives, just emulated ones. Therefore, you will end up with the page file and swap file on the emulated drive. Even if you aren't doing anything "interesting" with your machine, the pagefile and swapfile will still see activity. If you start doing things that need lots of memory, starting VMS Mail, or running a compiler, or using a word processor like MASS-11 (to name a few things we used to do all day long with our departmental VAX 15+ years ago), there will be a flurry of reads and writes at each image launch. While it would "work" to run a VAX from a CF drive, I don't think the lifetime of the CF card would be particularly high. I can't quantify it off the top of my head, but to speculate wildly, I think I'd rather depend on an RD53 drive as unreliable as they are (I still have several from the 1980s,. but I don't tend to use them as boot volumes). As for experience with Flash wear and shrinking capacity, I ran an "Earthdial" at the South Pole over sunny parts of 2004 and 2005. My camera was an old 2megapixel Olympus, chosen because a) it was in the cupboard, and b) it had a serial interface. I wrote a script to take a picture every 2.5 minutes, dump it, add some text at the top and bottom, and post it to my web page. Over a few weeks of operation, that flash card lost two photographs-worth of capacity from being asked to make 576 writes of a few hundred sectors, day after day. Percentage-wise, it was probably 3% of the capacity lost over 4 months of constant use. While the card would then last for years in that arrangement, it does seem like a fast decline to me. This card was an older one. Perhaps newer ones have 10 or even 100 times as many write cycles as ones ten years ago, but that only delays the inevitable. If Flash advances to the point that using it in a write-intensive environment would have a lifetime measured in multiple years, then it's probably a viable replacement for rotating media. We are, after all, accustomed to replacing hard drives every three to five years anyway. I don't think I'd want to intentionally set up a system where I have to be concerned about how fast it's degrading, and watch the hour-meter tick away its remaining lifetime. Yes, the argument could be made that disks do that anyway, but I don't think it's the same when you can set your calendar by the self-destruct date. -ethan