[2097] in Commercialization & Privatization of the Internet
Re: NREN and the Economics of Mainframes Versus Workstations
daemon@ATHENA.MIT.EDU (Einar Stefferud)
Sat Jan 25 22:31:17 1992
To: "David J. Farber" <farber@central.cis.upenn.edu>
Cc: com-priv@psi.com
In-Reply-To: Your message of Sat, 25 Jan 92 17:09:32 -0500.
Reply-To: Stef@nma.com
From: Einar Stefferud <Stef@nma.com>
Date: Sat, 25 Jan 92 19:21:18 PST
There is more to this Central CC (Mainframes or not) vs Local
(Workstations or not). The keys are in governance questions like
"accountability" and "responsibility".
If a Central CC gets a pot of money (taxes, subsidy, or budget grant)
from on high, and if it does not involve its customer/users in its
governance, then it becomes a competitor of its customer/users and
they cease to cooperate in their mutual best interests because they
don't have any mutual best interests.
I call this the "Authority Distribution Model".
In this model, the University subunits generally perceive that if they
can lobby against funding for the Central CC, in favor of another
faculty position, or some facilities of their own, they get more than
they give up, in terms of controllable resources, especially if they
value a faculty position more than some undivided share of Central CC
computing services. (They can easily think that they can get the
additional faculty position and also finagle to keep their current
consumption of CC services.) Every chunk of money going to the
Central CC is seen as money each subunit could have gotten for itself.
The service provider must compete with its users for funding!)
Why should anyone put their funds into an undivided share of a service
facility that is controlled by others?
There is another model which I call the "Power Aggregation Model",
which I have helped to install in a few universities over the years.
It involves only minimal direct support to the Central CC, with the
subunit "customers" buying services for their "users" to consume. The
subunit's computing funds may be partially restricted to purchases
from the Central CC, but must not be entirely so restricted, else this
model reverts back to being the Authority Distribution Model.
The subunits must have significant unrestricted funds to spend on
computing (or whatever they wish), and the Central CC must be set up
to be dependent on getting the subunits to spend their discretionary
funds on computing in return for perceived value of computing services
delivered. The CC may be subsidized, but only at minimal levels.
Then, to deal with the governance question, the central CC should be
placed under the control of a Board of Directors that is populated
with budget responsible representatives of the subunits which are the
customer base. The board gets to ask about and find out about any
aspect of operations or funding, and in essence operates like a
Customer Coop Board of Directors. Hence the name "Power Aggregation
Model" because it aggregates the real power of the university subunit
budgets into the oversight body of the Central CC management structure.
There is a major key to all this: A CUSTOMER is not a USER!
They may both use the same words, and often sound very much alike, but
they are not alike, do not mean the same thing when they use the same
words. They have almost unrelated (orthogonal) value systems.
The only thing they have in common is the intersection of their two
orthogonal value system hyperplanes -- for those of you who are into
image displays. Hyperplaning in the marketplace is such fun!
USERS are an infinite sink for resources. Given the opportunity to
use more, they will use more, and then ask for still more. USERS have
little or no incentive (other than altruism) to make value choices
which reduce their levels of resource consumption. (Just ask any
INTERNET USER, myself included!)
A CUSTOMER makes value judgements. When a customer asks for
something, it is an offer to buy and pay for it.
So, USERS run up bills by using services, and CUSTOMERS decide the use
is worth the cost and decide to pay the bills, or decide the use is
not worth it and cut back on purchases.
So, ask the obvious questions in any given situation:
Who are the customers (the ones who decide what to pay for) and who
are the USERS (the ones who consume services and run up the bills)?
In Central CC situations, the CUSTOMER is the President of the
University if the CC gets its budget from on high. In this case, CC
management soon learns that what the President really wants is to not
hear about any problems, and the CC management becomes very adept at
greasing squeaky wheels and paying much attention to campus politics.
How else can the CC management keep the Presidential Customer happy?
Or, if the CUSTOMERS are the subunits, then the money flows through
them to get to the Central CC. Guess who gets attention now? Squeaky
wheels with no money are sent off to argue with the Budget Manager and
not with the CC Manager. (Pretty neat, Eh?)
Now then, overlay this fabric with technology advances that provide
continually dramatic unit cost decreases for computer facilities (PC
and Workstation systems), and the result is entirely predictable.
Guess where it is going to be in anther 10 years! Pretty easy to use
this crystal ball, I'd say. (Hint: Its called distributed systems.)
Now, we got into this bit of discussion from our discussion of NSFnet
and COM-PRIV. Seems to me that all this is quite applicable, and I
think Steve Wolff's long term plan is well aligned with the
projections I have implied here. He is trying over time to convert
the NSFnet USERS into CUSTOMERS in a world that is too full of
Authority Distribution governing systems. (Few if any CUSTOMERS!)
In this situation it is very hard (within the constraints of US
Government Rules -- A Great Authority Distribution System) to shift
from direct NSFnet facility/service purchasing to University Funding
(another Great Authority Distribution System for Central Facilities)
or Principal Investigator Funding (Power Aggregation at last) such
that Universities and Principal Investigators will actually continue
to buy the services offered now by the current NSFnet providers.
Continuity is a critical requirement, and not easily put at risk.
I expect that it would really help if there was a substantial private
sector competitive market alternative already in place. (Obviously it
is not in place, as I read COM-PRIV;-).
The problem before us is how to get it to fall into place. I don't
think it makes much sense to beat on NSF for not having already put it
in place, cause that is just not something the NSF can (or should) do.
Unfortunately, we don't seem to have an over supply of non-NSF people
who know how to do it either!
Onward!...\Stef
