18 May
2005
18 May
'05
12:08 p.m.
On 5/18/05, Rob O'Donnell <classiccmp.org at irrelevant.fsnet.co.uk> wrote:
The worst processor/user ratio I think I encountered
was about 50 users
sharing a 486dx50.. It was in a rack case, in the top 6" of a 4' rack, the
rest being blocked in empty space except for a UPS sat in the bottom. The
users most definitely called it "the mainframe" !
Worst processor/user ratio for PeeCee-class equipment, or on a
per-cycle basis? We used to routinely put 40-60 users on our
VAX-11/750 (0.6 VUPs, 8MB of RAM, *roughly* the same CPU horsepower as
an 8MHz 68000, but with broad I/O channels). When more than 50% of
our users were active, memory was a problem. We used to have to
encourage people not to sit in MAIL or MASS-11 (word processor app)
because even with shared code segments (each user maintained their own
swappable store), more than about a dozen users sitting idle in those
apps would absorb most of the free memory.
We used that machine for intra-company e-mail, sales letters, and
software development (in-house assemblers, Whitesmith's C, VAX-C...),
and product testing.
_That_ was by all standards of the day, a mini, not a micro. Later
machines had much faster processors, and somewhat more memory, but the
I/O architectures of those later machines were much more sparse.
Three 32-bit I/O channels (MASSBUS), and a 16-bit asynch I/O bus
(Unibus) trumps much of the early PC era substantially.
So in the great mini-vs-micro debate, once one is talking about later
16 and 32-bit minis (early 12 and 16-bit minis do tend to have one
medium-performance I/O bus), I'd have to say that I/O architecture has
as much to do with the definition as the number of processors.
-ethan