mac3n@babbage.acc.virginia.edu (Alex Colvin) (06/07/88)
> Third-party operating systems that replaced the manufacturer's system: > > GE's CTSS runs on the Honeywell DPS-8. Uh... That's Dartmouth's DTSS, or GE's Mark <some roman numeral> that run on the Honeywell (formerly GE) mainframes (formerly called 635/66xxx, now DPS8). CP-6, descended from the Xerox OS, also runs. Multics runs (ran) on its own offshoot of this architecture. That's a lotta OSs, considering that GCOS used to come bundled with the hardware.
davidsen@steinmetz.ge.com (William E. Davidsen Jr) (06/10/88)
In article <321@babbage.acc.virginia.edu> mac3n@babbage.acc.virginia.edu (Alex Colvin) writes: | Uh... That's Dartmouth's DTSS, or GE's Mark <some roman numeral> that DTSS is (was) interesting in that it is one of the systems which allows only one processor to handle interrupts. I believe that all of the other systems for this cpu use all processors for all functions. We were visiting Dartmouth back in ~1969 to see this wonder work. One of the grad students was showing us around and was talking about how reliable the system was if something failed. He walked over to one processor and just pushed the "power off" button. All you heard was the console typing a message that the cpu was gone. One of our group asked what would happen if that had been the control processor. The student walked over to the control processor (after bringing the original processor back online) and punched *its* power off. The tapes stopped spinning. The printers stopped printing. Not a light blinked anywhere in the room. Then in the distance came the chatter of one console printing something like "service timeout on master cpu, deconfigured. Processor C is now master." Then everything started again. While we were there we looked at the "RACE file." This was a monster box full of small Mylar cards. On demand a card would be selected and slid down to the business end, where it would be wrapped around a drum and read as a small drum device. I believe the transfer rate was 72 bits every 7us, or about 10 meg. This was screaming in the days of 200ms access disks. Later GE built the 270 disk, a head per track job. They cost a bunch but were really fast. -- bill davidsen (wedu@ge-crd.arpa) {uunet | philabs | seismo}!steinmetz!crdos1!davidsen "Stupidity, like virtue, is its own reward" -me
haynes@ucscc.UCSC.EDU (99700000) (06/10/88)
In article <11202@steinmetz.ge.com> davidsen@crdos1.UUCP (bill davidsen) writes: >In article <321@babbage.acc.virginia.edu> mac3n@babbage.acc.virginia.edu (Alex Colvin) writes: > >| Uh... That's Dartmouth's DTSS, or GE's Mark <some roman numeral> that > > DTSS is (was) interesting in that it is one of the systems which >allows only one processor to handle interrupts. I believe that all of >the other systems for this cpu use all processors for all functions. > I don't know about the later Honeywell machines, but on the GE-635 it is the hardware, not the software, that requires one processor to handle all the interrupts. I have a mental picture of a rotary switch, but my memory isn't very good. The Burroughs B5500 also had this "feature", in this case controlled by a toggle switch that designated which of two processors was to handle the interrupts. The Burroughs D-825 (military machine, different division of Burroughs) supposedly did let any processor handle the interrupts so you really could just switch off any unit and the system would continue operating. On the B5500 you could switch off any I/O channel if you could catch it between operations. Or you could switch off any I/O channels so long as you left at least one, and you could switch off any memory except the one at address zero, and you could switch the second processor on or off, and then reboot and you were running again. Which was quite a boon contrasted with some competing machines that were totally down when any part was down. > We were visiting Dartmouth back in ~1969 to see this wonder work. One > ... > While we were there we looked at the "RACE file." This was a monster >box full of small Mylar cards. On demand a card would be selected and >slid down to the business end, where it would be wrapped around a drum >and read as a small drum device. I believe the transfer rate was 72 bits >every 7us, or about 10 meg. Gosh, I didn't realize G.E. ever let one of those things out of the plant. The drive was actually made by RCA, and inspired such confidence in all who saw it that it was universally called the "chicken plucker". > This was screaming in the days of 200ms >access disks. Later GE built the 270 disk, a head per track job. They >cost a bunch but were really fast. Actually a Burroughs disk OEMed to GE. GE's previous disk line was one of those things with 3-ft diameter horizontal disks, each with its own separate head actuator. This was a product originally developed for the GE 200 computer line, then ported to the 400-line where it was almost adequate, and then further ported to the 600-line which beat it to pieces. >-- > bill davidsen (wedu@ge-crd.arpa) > {uunet | philabs | seismo}!steinmetz!crdos1!davidsen >"Stupidity, like virtue, is its own reward" -me haynes@ucscc.ucsc.edu haynes@ucscc.bitnet ..ucbvax!ucscc!haynes
davidsen@steinmetz.ge.com (William E. Davidsen Jr) (06/13/88)
In article <3707@saturn.ucsc.edu> haynes@ucscc.UCSC.EDU (Jim Haynes) writes: | I don't know about the later Honeywell machines, but on the GE-635 it is | the hardware, not the software, that requires one processor to handle all | the interrupts. ... Yes, I certainly said that one wrong. What I meant was DTss has only one processor in master mode, whereas GECOS had all processors running around in the o/s. This actually caused very little problem unless someone forgot to close (or worse open) a gate on critical code. | > This was screaming in the days of 200ms | >access disks. Later GE built the 270 disk, a head per track job. They | >cost a bunch but were really fast. | | Actually a Burroughs disk OEMed to GE. GE's previous disk line was one | of those things with 3-ft diameter horizontal disks, each with its own | separate head actuator. This was a product originally developed for the | GE 200 computer line, then ported to the 400-line where it was almost | adequate, and then further ported to the 600-line which beat it to pieces. I believe that the numbers were DSU110 for the "small" one and DSU160 for the "big" one, which held 16MW (64MB). It was as big as a Volkswagon, and had the heads run by linear motors (if I recall correctly, the only two other survivors of those days are out today). -- bill davidsen (wedu@ge-crd.arpa) {uunet | philabs | seismo}!steinmetz!crdos1!davidsen "Stupidity, like virtue, is its own reward" -me