home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!auspex-gw!guy
- From: guy@Auspex.COM (Guy Harris)
- Newsgroups: comp.arch
- Subject: Re: multiuser systems (was Re: IBM AS/400 is the world's slowest..)
- Message-ID: <16189@auspex-gw.auspex.com>
- Date: 2 Jan 93 04:25:55 GMT
- References: <id.TX9W.FC3@ferranti.com> <1993Jan1.102554.28575@metapro.DIALix.oz.au> <C07DK0.E7t@cs.bham.ac.uk>
- Sender: news@auspex-gw.auspex.com
- Organization: Auspex Systems, Santa Clara
- Lines: 55
- Nntp-Posting-Host: auspex.auspex.com
-
- >People who suffered timeshared systems in the 70's often refuse to
- >believe any of this unless they have direct experience of the
- >benefits.
-
- I suspect the people who suffered timeshared systems back then suffered
- from having a timesharing system capable of supporting N people
- comfortably, but having M > N (perhaps M >> N) people using it
- regularly.
-
- Is it that at least some of the resources that were in too scarce supply
- back then - CPU and memory - are now in much more plentiful supply, so
- that you're less likely to have an overloaded time-sharing system?
-
- (Yes, I remember early-'80s timesharing systems. I've *no* desire to
- use a timeshared system that can't, for example, give quick response to
- keystrokes in editors such as Frame Maker, even if other folks are doing
- compiles, running their own sessions with Frame or 1-2-3 or whatever.)
-
- >The strength of this argument is not undermined by the horrors of X:
- >at least it provides graphical interaction over a network.
-
- And not all such systems (central compute and file servers, desktop
- graphics machines that aren't expected to do all the compute work
- themselves) use X; see, for example, Bell Labs's Plan 9.
-
- Rob Pike has made many of the same arguments (lower cost for desktop
- machines, ability to upgrade the central machine's CPU power instead of
- upgrading all the desktop machines) for the Plan 9 system (although,
- Rob's tendency, at least at one point, to refer to the desktop machines
- as "terminals" nonwithstanding, they're actually diskless workstations,
- running essentially the same OS as the compute servers, and capable of
- running applications, although they generally have less CPU power and
- probably less memory than the compute servers).
-
- The Plan 9 model is actually a bit different from the "dumb terminal"
- model, even the "dumb X terminal" model, as applications that don't
- consume *huge* amounts of CPU, and that are quite interactive - e.g.,
- text editors (even ones such as Frame; they probably consume more CPU
- than "vi", but probably not as much as Verilog...), spreadsheets, blah
- blah blah, can be run on the desktop machine.
-
- Applications requiring more CPU or memory, or requiring a tighter
- connection to the file server (the compute servers and file servers are
- joined by higher-speed links than are the desktop computers and file
- servers), but not requiring quick response to keystrokes or mouse
- motions are probably run on the compute server. (Graphical applications
- that talk directly to a window, and get keyboard and mouse events
- directly, can be run either on the compute server or the desktop
- computer; the window system is one that runs over a network.)
-
- (If you want to debate the merits of Plan 9, debate them with Rob Pike
- or Ken Thompson, not with me. I've never used it, I've just heard about
- it and read some papers on it. I'm pretty much just reciting its
- creator's views, as I understand them; I've not enough information to
- agree *or* disagree with them, at this point....)
-