In article <61267@bbn.BBN.COM> cosell@bbn.com (Bernie Cosell) writes: > ... The 'card' referred to was a >*magnetic* card. I forget the exact size but it was something like 3" >wide by maybe 8" high. It held some modest [by modern standards], but >actually by back-then standards, reasonable amount of data. The >operation was kind of a kludge -- the more you looked at the >'principles of operation' book, the more you wer eSURE that it was >designed by Rube Goldberg to get some kind of revenge on the computer >industry. This device reminds me of a couple of IBMs attempts at mass storage, the Data Cell and the MSS (Mass Storage [sub?]System). This place had both in its checkered history of computing. I arrived here too late to see the Data Cell in operation. I understand from doing some reading that it was a carrousel(sp?) sort of arrangement with long pie shaped widgets that hung on a frame of some sort. The pie shaped widgets had magnetic tape strips in them, I think fan folded, but I could be mistaken there. When you wanted to fetch data from the Data Cell the thing would find the proper widget and snaffle in the tape, I guess. Access times, if i recall correctly from my long since mislaid System/360 Configuration Guide, were measured in tens of seconds. Anyone out there who can correct my myriad misconceptions and maybe lay a little folklore about these beasts on us? The MSS was a slightly different bird. It was a long cabinet, maybe twelve or so feet long in its smaller configuration. Inside the cabinet, on both inside walls, was a honeycomb affair. The cells of this honeycomb would be loaded with little bullet shaped cartridges, about three or so inches long and about an inch and a half in diameter. The cartridge had a plastic sheath around it and inside the sheath was a spool with several feet of magnetic strip. The strip was about two or two and a half inches wide. There were two robot hands that ran horizontally on tracks and vertically on poles that would pick a cartridge out of the honeycomb, take it to an available read/write station, called a DRD for some reason, and stuff it in. The DRD would pop the sheath off the cartridge, grab the tape strip, and pull it past a fairly clever helical scan tape head. Fun came when the DRD would refuse to disgorge a cartridge after it was done. The rest of the system might be under the impression that the DRD was free for use and the picker would try to stuff a new cartridge into an occupied DRD. On occasion, the picker hand would get confused about its exact physical whereabouts and have to go off to a corner and sulk for a while. A fairly neat gizmo. The cartridges held about half of an IBM 3330 disk -- er -- excuse me -- DASD volume. Don't remember off hand how big those were, but I seem to recall about 150 MBytes or so. Anyone with a better memory than mine? (silly question) spl (the p stands for put my MSS cartridge WHERE?) -- Steve Lamont, SciViGuy -- 1882p@cc.nps.navy.mil -- a guest on network.ucsd.edu NPS Confuser Center / Code 51 / Naval Postgraduate School / Monterey, CA 93943 What is truth and what is fable, where is Ruth and where is Mabel? - Director/producer John Amiel, heard on NPR dwithers@ncratl.Atlanta.NCR.COM (Dave Witherspoon) writes: }I have heard that long ago (and well before I got out of school and }began working for NCR), NCR developed a machine called the "CRAM". Not }See-RAM...CRAM. Stood for Card Random Access Memory.... It certainly did! }Apparently, this }device was a huge machine that manipulated a horizontal stack of ordered }Hollerith cards. A program would request a particular card from the }"file". The CRAM would estimate the position of the card, select a }bunch +- around where the card should be. It would then }scan the bunch until it located the desired card, and return the data. }Clearly, you would open the "file" for "read-only" access :-). My how you conjecture... I susupect you've really invented this or you've been seriously misinformed. The 'card' referred to was a *magnetic* card. I forget the exact size but it was something like 3" wide by maybe 8" high. It held some modest [by modern standards], but actually by back-then standards, reasonable amount of data. The operation was kind of a kludge -- the more you looked at the 'principles of operation' book, the more you wer eSURE that it was designed by Rube Goldberg to get some kind of revenge on the computer industry. There was a drum, and the actually reading and writing happened when one of the cards was wrapped around the drug and then being handled as if it were a small magetic drum [drums were pretty popular, and moderately well understood, forms of bulk memory back then]. Mostly the drug grabbed one [short] edge of the card and spun, and the rest of the card just wrapped around. The fun part: handling the cards. Each card has 'punch holes' at one end [not like computer-card punch holes, but [if I rmember right] more like real 'three hole punch' like holes. There were reserved places for holes, and you, basically, binary-encoded the 'number' of a card by punching hoes in the right place. The cards were supported on some kind of rack, and when the computer requested a card, the first thing that happened was that the drum let go of its card, and there was some kludge for sucking-up the now-free card, and whipping it off to the holding area where it was added to the pack. Then the selected card # activated some kind of selection kludge that would (theoretically) allow ONLY the card with _just_ the right punched-id-number to fall free from the supporting bars. It would drop down into a funnel-like affair, from which it would be sucked over to the drum, which would grab the end of the card, and whip it around the drum: and voila you are randomly-accessing a new card! }They say that when a CRAM got into a bad mood, it could rifle cards all }over a computer room in no time at all.... I don't remember this as a failure mode, although I figured out right away that this was a beast to stay away from, and so I tactfully found a project to work on that kept me safely far away from the CRAMs. BUT: I do know that its primary failure mode was to jam. unbelievably. Gobbling lots of cards in the process [thereby losing BIG chunks of data]. I seem to recall that the _primary_ problem was that the card-drop-selector wasn't so wonderful [these were just sheets of magnetic stuff, no stronger than the stuff from which floppies are made, and the wear-and-tear on the cards make the support ears and the various selection-punch-holes be less and less reliable... until an improperly stacked card would be selected with the wrong ID #... and that means that [at least] two cards would drop... and it'd be crunch-central. Another problem was whippin ght e card on and off the drum, and getting it added to the stack of held-cards: just as photocopies regularly jam because zillions of variables make moving flexible, real-world things around in a hurry a less-than-deterministic activity. A card might jam in the return slot, and then [until the machine gave up], EVERY card coming off of the drum would micro-pack behind it as it unwrapped from the drum. [Jon: you out there? Can you fill us in on some CRAM anecdotes from the Telcomp days?] /Bernie\ moon ma In article <4143@network.ucsd.edu> slamont@network.ucsd.edu (Steve Lamont) writes: >This device reminds me of a couple of IBMs attempts at mass storage, the Data >Cell and the MSS (Mass Storage [sub?]System). This place had both in its >checkered history of computing. I arrived here too late to see the Data Cell >in operation. I understand from doing some reading that it was a >carrousel(sp?) sort of arrangement with long pie shaped widgets that hung on a >frame of some sort. The pie shaped widgets had magnetic tape strips in them, >I think fan folded, but I could be mistaken there. When you wanted to fetch I believe the IBM number was 2321 data cell. A popular configuration had a 2321 and a 2311 (25 mbyte hard disk) on one controller. This combo was often used for ISAM files where the 2311 held the index and the 2321 (aka noodle grabber) held the data. No fan folds here, the strips were flat, about 2 inches wide, 14 inches long, and held (and now my memory is fuzzy) 100 tracks of 2000 bytes each. >data from the Data Cell the thing would find the proper widget and snaffle in >the tape, I guess. Access times, if i recall correctly from my long since >mislaid System/360 Configuration Guide, were measured in tens of seconds. >Anyone out there who can correct my myriad misconceptions and maybe lay a >little folklore about these beasts on us? There were 10 of the pie shaped bins, each with 20 cells, each with 10 strips (again memory is fuzzy). To this day, IBM dasd is addressed as MBBCCHHR, where M was the cell (now represents the extent), BB the bin, CC the strip (now the cylinder), HH the strip (now the head), and R the track (now the head). Access time was in tenths, (not tens) of seconds. (I am sure it is MBBCCHHR, but the meanings are fuzzy and may be wrong.) The mechanism picked a strip and wrapped it around a drum to read/write it. An aquaintance told me of doing an experiment with the 2321. He allocated the 3 assembler work files, the punch file and the text file to bins 1, 3, 5,.. of the 2321 and started to assemble [a very large assembly]. After several hours (it normally ran about 1/2 hour) he gave up and killed it. >The MSS was a slightly different bird. It was a long cabinet, maybe twelve or >so feet long in its smaller configuration. Inside the cabinet, on both inside >walls, was a honeycomb affair. The cells of this honeycomb would be loaded >with little bullet shaped cartridges, about three or so inches long and about >an inch and a half in diameter. The cartridge had a plastic sheath around it I seem to rember the dimensions being a bit larger, about 1.5 times yours. >and inside the sheath was a spool with several feet of magnetic strip. The >strip was about two or two and a half inches wide. > >There were two robot hands that ran horizontally on tracks and vertically on >poles that would pick a cartridge out of the honeycomb, take it to an >available read/write station, called a DRD for some reason, and stuff it in. I seem to (fuzzily) remember the DRD was the Data Reading Device. >The DRD would pop the sheath off the cartridge, grab the tape strip, and pull >it past a fairly clever helical scan tape head. The helical scan was demo'd by IBM salesmen using their diagonally striped ties. :-) >Fun came when the DRD would refuse to disgorge a cartridge after it was done. >The rest of the system might be under the impression that the DRD was free for >use and the picker would try to stuff a new cartridge into an occupied DRD. > >On occasion, the picker hand would get confused about its exact physical >whereabouts and have to go off to a corner and sulk for a while. > >A fairly neat gizmo. The cartridges held about half of an IBM 3330 disk -- er >-- excuse me -- DASD volume. Don't remember off hand how big those were, but >I seem to recall about 150 MBytes or so. Anyone with a better memory than >mine? (silly question) Your description matches fairly well with my fuzzy memory. 3330 disks came in two models, model 1 was 100 mbytes, model 11 was 200 mbytes. The MSS cart held 50 mb and filled 1/2 of a m1. Rich This is some filler so rn wont reject my post. The opinions above are mine, you can't have them. I last used a data cell about 1968, a data cell about 1973. (I have been in this business too long). In article <20201@.la.locus.com> richg@locus.com (Rich Greenberg) writes: >In article <4143@network.ucsd.edu> slamont@network.ucsd.edu (Steve Lamont) writes: >>The MSS was a slightly different bird. It was a long cabinet, maybe twelve or >>so feet long in its smaller configuration. Inside the cabinet, on both inside >>walls, was a honeycomb affair. The cells of this honeycomb would be loaded >>with little bullet shaped cartridges, about three or so inches long and about >>an inch and a half in diameter. The cartridge had a plastic sheath around it > >I seem to rember the dimensions being a bit larger, about 1.5 times yours. I just found an old one and measured, if anyone really cares :-): They were 1 3/4 inches in diameter, 3 1/2 inches long, and the tape was 2 3/4 inches wide. Capacity was 50 megabytes, or half of a 3330V (Virtual 3330) disk. According to the systems guy I just asked, he sez that they were always mounted in pairs. spl (the p stands for probably been in this business too long, myself...) -- Steve Lamont, SciViGuy -- 1882p@cc.nps.navy.mil -- a guest on network.ucsd.edu NPS Confuser Center / Code 51 / Naval Postgraduate School / Monterey, CA 93943 What is truth and what is fable, where is Ruth and where is Mabel? - Director/producer John Amiel, heard on NPR In article <20201@.la.locus.com> richg@locus.com (Rich Greenberg) writes: >In article <4143@network.ucsd.edu> slamont@network.ucsd.edu (Steve Lamont) writes: [details of 2321 data cell deleted for sake of sanity...] And the VTOC was on the first strip in each of the 10 cells. That strip got plucked far more times than any other strip in each of the cells. And wore out. And, if my memory is correct, you had to run stand-alone to DASDI a CELL until somebody figured out that if you used the cell much, you'd spend all your time stand-alone so you couldn't use the cell much. If you used IEBGENER to copy a dataset from one cell to another cell, (say from cell01 to cell06, on opposite sides of the washtub like mechanism), the drill was 1) pluck the strip from cell01, 2) wrap the strip around the little drum so it could be read, 3) read the strip, 4) unwrap the strip from the little drum, 5) stuff the strip back into cell01, 6) rotate the bin 180 degrees, 7) pluck the strip from cell06, 8) wrap the strip around the little drum so it could be written, 9) write the strip, 10) unwrap the strip from the little drum, 11) stuff the strip back into cell06, 12) continue 1-11 until you get sick watching the mechanism... The "correct" way to copy a dataset from one place to another in the cell was to use two job steps and a temporary on one of your many 2301's :-) I think the original design for a 2321 included a 2301 for a cache, but the drum was kissed off to save money. Or maybe to create a bigger market for the replacement magnetic strips... :-) -- Mike Murphy mrm@Sceard.COM ucsd!sceard!mrm +1 619 598 5874 This discussion of the Data Cell and the CRAM units reminds me of the RCA RACE (Random Access Card Equipment) we had at Carnegie-Mellon in the early to late 1960's when I was an undergraduate. The description given regarding the failure modes for the CRAM units were nearly identical to those of the RACE units. First this was an attempt to provide the Bendix/CDC G-21 (yes, Bendix) computers at Carnegie with mass online storage for the teleprocessing terminals (Teletypes). If I remember right, there were four RACE units in use, each of which contained 8 cassettes which contained approximately 100 magnetic cards. The amount of storage was fairly massive by 60s standards, potentially totally around 500 or so megabytes. But the units were so massively unreliable that all data was stored in duplicate on two units; and the additional two units served to back up the two working units. Effectively there was only around 100 megabytes or so. The cards were approximately the same size as those described for the Data Cell which I also saw, but we did not have such a unit and really never considered one given the experience with the RACE unit. The cards were around 4" by 12" and the selection of the card from the cartridge was by notches. Once the card was selected it was moved by either rollers up to four feet to a rather large drum containing the read-write heads. The card would wrap around the drum and a head would step up and down. The failure modes were always interesting and rather dramatic. One of the more common modes was simply that the card would crunch as it entered the read/write drum, resulting in a jam. As an operator, you could had one option - yank the card out similar to the approach used for a jammed sheet of paper in a copier. Obviously, the card was not very useable after that. Less often, two cards would be selected and result in a crunch. Most often was that the cards became unreadable due to the extensive wear and tear of the selection and reading process. I really do not know what the life of the cards were like, but on a weekly backup, which somehow I always seem to be the operator, we would lose 10 to 20 cards easily. It was also not uncommon, as a user, for me to lose my programs even though they were stored in duplicate. In reality, I wonder whether cards were ever replaced once the operators yanked them. The interface between the G-21 was through an RCA-301. Since the 301 interacted with two CPUs, there always seemed to be deadlocks which needed to be resolved. Usually this was done by reloading the 301 since the computers were usually hopelessly confused. The 301 was also not the most sophisticated machine and the interface must have been extremely primitive. The initialization sequence required an entry of 20-30 instructions on a front panel (which I got very good at) and this resulted in a boot off a 7-track tape drive. Oh well, those were the days. Dave Chou ncoast!dchou@usenet.ins.cwru.edu