OFFICE MEMORANDUM e@ STANFORD UNIVERSITY © OFFICE MEMORANDUM e STANFORD UNIVERSITY © GFFICE MEMORANDUM

To

From

Supsect:

Date: May 1, 1974

J. Lederberg/T. Rindfleisch

E. Levinthal Ko

Allocation of SUMEX Computer Resources

I would like to suggest the following method of handling the allocation
Problem as an alternative to use of a "currency".

We define a few classes of users (three may be sufficient). The membership
in the highest or primary priority class should be limited to primary
projects (not necessarily all projects) of users who are central to the
goals of SUMEX. The number in this class should initially be small enough
so that we can be certain this use by itself will lead to an acceptable load
factor. The secondary classes should include secondary users aid secondary
projects of primary users. Similarly the third class would be non-priority
users and non-priority projects of primary and secondary users.

A user signs on with a particular project name which has received a
classification. The user can ask for a histogram which would give graphs of
usage for each class by time of day and day of week. These graphs should

be designed so that both what was used and what was available can be deduced.
This should be intezrated back sufficiently far in time to give good

Statistics but not so far back that it doesn’t represent the current distribution

of users. Secondary and then tertiary users "get on" only if the load factor
is less than a predetermined maximum.

This system would have the following advantages:
1. “2 users could adjust their scheduled use of the most favorable

times without having to make detailed (and probably unsuccessful)
intra or inter-class predictions of use or negotiations for use.

2. It avoids a difficulty that existed with ACME “funny money" a.locations,
namely the reluctance to turn off users who "ran out" of ailocacions,

whose expected grants didn't get funded, etc., etc. "Delay is «
ceadly form of denial" but it is easy to administer. This problem
Giminished but didn't disappear in the case of Provost Funds.

«ne fact that the source of "funds" was remote from users nelped.
uowever, there still remained the problem of collecting overdrafts.

3. This system could and should also be used for system development

work. There will undoubtedly be a great deal of language improvements,
auditions of systems features, etc., that while nice are diversionary

from primary goals. By assigning such systems tasks secondary

or even tertiary status, efforts in these directions can be contained.

4. it still allows "currency" to be used for some resources such as
storage and perhaps I/O facilities.

ALISHU3SAINN GYOINVIS e

© WOONVYOWIW 3Dd1sIO e&

® ALISYJAINN GYOINVIS

@ ALISHZAINN GYOINVLIS © WNAONVYOWSW 3d1dd0

WNONVYOWIW JADIdIO