15 Jun
2005
15 Jun
'05
4:05 a.m.
Subject: Re: PDP-8s and -10s
From: Al Kossow <aek at bitsavers.org>
Date: Tue, 14 Jun 2005 20:50:02 -0700
To: classiccmp at classiccmp.org
It seems to me that they could have made
the -10 line real mainframes (no, IBM and the BUNCH did not consider the
-10 line real) with just a little effort and not much money.
--
DEC through the 'golden age' didn't go after that market. They tried to
target scientific rather than corporate computing.
One of the oddities is that the 36bit machines the 10s/20s were well
known for their timesharing with huge nubers of users.
One of the things that isn't talked about much is
how the PDP-6 just about
killed DEC because of all of the resources it consumed to get it out the door.
The PDP-6 was the VAX-9000 of the 60s.
They did eventually abandon their 'fan base' to go after the corporate
market, and that was the beginning of DEC's downfall.
It was successful for the 80s but the growing PC price performance
numbers were hurting the backroom mainframe style systems. The
breakpoint was when 486 powered system with basic decent graphics
(640x480x8) hit the streets with Win3.11. We had the scientific
workstation wars of the late 80s this would be the business
workstation war. Several PCs for 2500$ each or less was cheaper
than a microvax of any flavor.
Allison
15 Jun
15 Jun
8:06 a.m.
One of the oddities is that the 36bit machines the
10s/20s were well
known for their timesharing with huge nubers of users.
Also interesting to note that the /other/ great timesharing machines of the
pre-Unix* era were also 36-bit: the GE/Honeywell/Bull 600/6000/Level 66/68
running G(e)COS or Multics.
... but then the 32-bit machine came along well after timesharing had become
established and it was some time before the '360 family had a decent ability
in this field - and by the VAX era the tendancy was to have more systems
rather than more users/system.
* I wonder what is the largest number of concurrent "tty" users that is
reliably reported as having been run on a Unix system. I know that we
replaced one GCOS Honeywell dual-processor 66/60 with 3 Unix Gould 6000s to
manage about the same number of users (tho' they were increasing their
processing demands**) ... but the _purchase_ cost of the Goulds was less
than 3 years _maintenance_ for the Honeywell.
The joke in those days about an editor that we refused to make available was
"Eight Megabytes And Continually Swapping" - nowadays who notices 8
megabytes :-) [unless they're programming PICs!]
Andy
10:14 p.m.
On Wed, 15 Jun 2005, Andy Holt wrote:
* I wonder what is the largest number of concurrent
"tty" users that is
reliably reported as having been run on a Unix system. I know that we
replaced one GCOS Honeywell dual-processor 66/60 with 3 Unix Gould 6000s to
manage about the same number of users (tho' they were increasing their
processing demands**) ... but the _purchase_ cost of the Goulds was less
than 3 years _maintenance_ for the Honeywell.
It's a slight change of venue, but Phil Becker's TBBS (I think it
was) ran 64 lines on double-digit MHz Intel processors... full
utilization. Written in assembly language, his code didn't
survive the historical onslaught but it was pretty spectacular at
the time. Some website says there was an alleged custom 96-line
version!
This is 64 serial ports (in N multi-port cards) in a high-end (for
mid-1980's) PC chassis, not special hardware support.
I think Phil wrote space-probe software in a previous life.
http://www.tbbs.org/TBBS.html
16 Jun
16 Jun
12:26 a.m.
On Wed, 15 Jun 2005, Tom Jennings wrote:
On Wed, 15 Jun 2005, Andy Holt wrote:
> * I wonder what is the largest number of concurrent "tty" users that is
> reliably reported as having been run on a Unix system.
I have seen
installations with 300+ users online. Obviously, these
were ran with terminal multiplexers, terminal servers and other
kinds of smart front-ends, as to not kill the main system with
line interrupts :)
Several of these systems were HP9000 'K' class servers, running in
the (centralized) city- and county governments in Holland.
--f
7:17 a.m.
>>>> "Fred" == Fred N van Kempen
<waltje at pdp11.nl> writes:
Fred> On Wed, 15 Jun 2005, Tom Jennings wrote:
> On Wed, 15 Jun 2005, Andy Holt wrote:
>
> > * I wonder what is the largest number of concurrent "tty" users
> that is > reliably reported as having been run on a Unix system.
Fred> I
have seen installations with 300+ users online. Obviously,
Fred> these were ran with terminal multiplexers, terminal servers and
Fred> other kinds of smart front-ends, as to not kill the main system
Fred> with line interrupts :)
Not Unix but timesharing, and more demanding at that -- the University
of Illinois PLATO system had 1000 terminals and most days hit its
configured limit of 600 active users, in the 1970s, on a 10 MHz
processor.
Its modern day successor serves a lot more terminals than that; some
vague memory says a million connect hours per month...
paul
23 Jun
23 Jun
1:48 p.m.
>>>> "Paul" == Paul Koning
<pkoning at equallogic.com> writes:
>>>> "Fred" == Fred N van Kempen
<waltje at pdp11.nl> writes:
Fred> On Wed, 15 Jun 2005, Tom Jennings
wrote:
>> On Wed, 15 Jun 2005, Andy Holt wrote:
>>
>> > * I wonder what is the largest number of concurrent "tty" users
>> that is > reliably reported as having been run on a Unix system.
Fred> I have seen installations with 300+ users online. Obviously,
Fred> these were ran with terminal multiplexers, terminal servers and
Fred> other kinds of smart front-ends, as to not kill the main system
Fred> with line interrupts :)
Paul> Not Unix but timesharing, and more demanding at that -- the
Paul> University of Illinois PLATO system had 1000 terminals and most
Paul> days hit its configured limit of 600 active users, in the
Paul> 1970s, on a 10 MHz processor.
Paul> Its modern day successor serves a lot more terminals than that;
Paul> some vague memory says a million connect hours per month...
Just got this bit of stats about that system: peak number of active
users at any one time: 9037. That's on a SINGLE machine (an Alpha).
I suspect that's a record for number of interactive users served by a
single computer. Especially if you consider that these are
interactive (key by key interaction) terminals -- not local forms
editing terminals as in the classic airline reservation systems
(SABRE).
paul
7123
days inactive
7131
days old
5 comments
5 participants
participants (5)
-
ajp166＠bellatlantic.net -
andyh＠andyh-rayleigh.freeserve.co.uk -
pkoning＠equallogic.com -
tomj＠wps.com -
waltje＠pdp11.nl