[2218] in Commercialization & Privatization of the Internet
Re: Low end users
daemon@ATHENA.MIT.EDU (Richard Mandelbaum)
Fri Jan 31 13:15:05 1992
To: Marc Horowitz <marc@MIT.EDU>
Cc: perry@mcl.unisys.com (Dennis Perry), com-priv@psi.com
In-Reply-To: Your message of Fri, 31 Jan 92 01:24:17 -0500.
Date: Fri, 31 Jan 92 13:11:42 -0500
From: Richard Mandelbaum <rma@tsar.cc.rochester.edu>
Marc, I don't question that "distributed environment can be run without
significantly more people than a central computing facility"
All too often , of course, distributed computing supplements rather
than replaces central computing because there arre still some things
that at MIT or Rochester etc. are better done in a central facility.
The problem is that "one size fits all" is true neither in Womens
hosiery or in computing. All too often there are decision makers
in place who don't understand that.
I have heard people whom the world considers intelligent including
VP's of major corporations and Universities claim that
"distributed computing" (whatever that is) would save 90% of
their computing costs without ever knowing that even when one
does distributed computing one still needs consultants and relying
on BMUG or their equivalent to support corporate users or
University users in a production fashion is not realistic.
I think it is this type of muddle headeed thinking which really
bothers me and i assume Dennis.
____________________
>> with distributed systems there is a system manager for each machine
,
>> but he is not classfied as a systems manager and therefoe we don't
>> have one and are not paying for one. This is true of workstations
as
>> well. Each workstation owner is his own system manager.
This may be true in many cases, but it is not a universal law. At
MIT, we have over 1000 workstations on campus. Almost 400 are in
public areas. Even if you want to call the private ws owners de-facto
maintainers, the public machines clearly have no such distinguised
individual, and many of the private machines are in the same
configuration as the public ones. With no customizations, no
maintenance is necessary. All 1000 of these machines (and their
servers) are maintained by a group of about 10 people, with 24x7 pager
service. The network is run by about five people, if you want to
count them, but they have other tasks as well, and the
one-person-per-machine model ignores the net, anyway.
(My example also excludes consultants and developers, but that still
only accounts for another 50 people, plus part-time student
consultants.)
My point is, that a distributed environment can be run without
significantly more people than a central computing facility (yes, MIT
has one of those, too). And with passable unix workstations available
for a few thousand dollars, the incremental cost to support a new user
is minimal.
Marc
