Talk:Gerrit Blaauw
Appearance
![]() | This article is rated Stub-class on Wikipedia's content assessment scale. It is of interest to the following WikiProjects: | ||||||||||||||||||||
|
First practical implementation of virtual memory?
[edit]What rendered the 360/67 the first practical implementation of virtual memory, as opposed to, say, the GE 645 on which Multics ran, or, for that matter, the segmented-but-not-paged B5000 machines or the paged-but-not-segmented Atlas at the University of Manchester? Guy Harris 22:09, 28 November 2006 (UTC)
- You raise a good point. I have basically been parroting this claim, which probably needs to be reconsidered. My read of the timeline is as follows.
- The ATLAS indeed had virtual memory but it suffered from severe usability problems. These were eventually diagnosed as due to thrashing. Hence the phrase "practical implementation". Period and retrospective literature I see discussing the system seem to regard ATLAS paging as not yet really practical.
- Virtual memory on the GE 645 certainly was practical by the time MULTICS was deployed, circa 1969. The question is whether it was "practical" on the 645 in 1966, when the S/360-67 was shipped, or more importantly in 1967-68, when CP-40 and CP-67 were up and running.
- I hadn't really thought about the Burroughs machines in this context. Perhaps the phrase should be first commercially practical implementation of paged virtual memory? I have always been a fan of both the ATLAS and the ALGOL machines; and of course MULTICS stands alone. On the other hand, I don't think there were many stable paged virtual memory systems in the 66-68 timeframe other than the CP/CMS systems. Trevor Hanson 04:00, 29 November 2006 (UTC)
- Having thought further about this, I believe that these other important systems have been under-described here (and elsewhere). There seems to be no question that CP/CMS broke new ground in terms of full virtualization; and I believe that CP/CMS, VM/370, and especially VP/CSS set the bar for contemporary time-sharing performance in terms of number of users per hardware dollar by year – though I can't find anything concrete about precisely when Multics achieved what kind of volume. However, all these engineers were "standing on each others' shoulders", and important work was done in many places. I've been focusing on getting the CP/CMS story straight, but it would be great if you or others could try to insert more details from the non-IBM world. Unfortunately, good sources are hard to locate. (My wife worked at CDC in those days, and I had a former business partner who was at Burroughs, and another who was a Multician; so I've heard countless anecdotes; but there seems precious little currently in print.) Trevor Hanson 06:24, 8 December 2006 (UTC)