OFFIC” MEMORANDUM e¢ STANFORD UNIVERSITY @ OFFICE MEMORANDUM © STANFORD UNIVERSITY e OFFICE MEMORANDUM

To

From

SUBJECT:

40072

DATE: April 9, 1974

File
Joshua Lederberg

Management Policies of the NCAR Computer Facility, another disciplinary resource.

I had a good talk today with Joe Oligar who is the Director of the Computer
Facility at NCAR, This is an inter-university consortium which is funded by
the National Science Foundation, It has a scientific staff of about 125
people and includes a rather comprehensive computer facility, some big CDC
machines, which are oriented to very large batch programs only,

They have a Board of Directors which represents the university consortium
as far as the computer allocation is concerned. Their main responsibility

is simply to make the first allocation as between time allowed for insiders,
outsiders and joint projects,

The insider usage is managed through internal procedures,

They have an advisory panel of computer science people and some atmospheric
scientists from the participating universities separate from the overall
board. This panel sets policies for the outside segment and it runs pretty
much as you might expect. I am hoping to receive from Oligar some documents
which will give more formal details about their procedures including their
algorithm for counting computer resource units.

They allow loal approval without going through the advisory board for usages
up to 20 or 50 hours of 6,600 machine time. All larger usages, plus all usages
of any side from people within 50 miles of NCAR must be approved by the
advisory panel, although they are able to make considerable delegation of

this responsibility, They expected a good deal of trouble and had just a
little with respect to conflicts of purpose from people at Boulder and Denver,
The main problem evidently was competition with their local university
computer centers, especially at Fort Collins. So, all of the local people are
given the additional barrier of having to go through the advisory panel. As
far as other outside users are concerned, the threshhold for informal approval
is about 1% of the annual resource. The advisory panel is named by one of

the deputies to the director of NCAR. Presumably this entire procedure was
authenticated by the overall board of directors.

The advisory panel meets 2 or 3 times a year and they generally require about

6 months turnaround time! However, it should be kept in mind that they are

dealing with huge projects involving very significant parts of a very large
computer resource, For smaller number of hours provisional approval can be

given locally so that people can be doing some preliminary work and debugging
before they have had formal advisory board approval for their big projects.

They can also do occasional mail votes which take two or three months to turn
around, They look very closely at the scientific value of the projects and

also at the computer technical efficiency of the algorithms that are proposed.

This obviously requires a great deal of work on the part of their review committee.

WNANVEOWSW 391ddO © ALISYTAINN GYOINVLIS © WNGNVYOWAW 351ddO © ALISYZAINN QYOINVLS © WNGNVYOW3W 3ddO © ALISYTAINN GYOINVIS ©
File -2- 4/9/74

They will send out proposals to other confidential consultants for further
advice. The content but not the names associated with these criticisms
are returned to the applicants.

The NCAR group has reserved the right to refuse projects regardless of
agency funding and presumably vice versa.

They provide no programming support at the present time for outside
users although they have a substantial staff to support the inside work,
and of course a program library is a by-product of this.

As part of the application forms they generally require an endorsement
from the computer center of the applicant's institution, and they also
notify those centers of approved programs. All this presumably is to take
care of the unfair competition problem.

Their resource is completely subsidized and is allocated on a no-cost
basis. However, they are starting to consider the need to charge for
expendable supplies.

They think it is very important that people outside have a clear idea
as to the length of leetime that will be required to process their
applications and they have had many examples of unreasonable and un-
fulfillable expectations.

They also run a summer program involving perhaps 10 graduate students to
introduce them to the applications of computing in atmospheric physics.

Oligar was interested in how we were proposing to run our program and
I promised to get him some documentation eventually, namely when we have
our own.

At NCAR they have not had any great pressure to institute even fairly
elementary time-sharing service such we enjoy at Stanford and of course

the pressures on the systems would be greatly enlarged if they did anything
other than big batch processing. They are expecting to introduce remote

job entry in the near future and Oligar will be very interested to see

how they can manage from there on.

cc: Tom Rindfleisch
Elliott Levinthal