On the need for human rationing (abstract only)

The occasion for this discussion is our recent experience with a severe shortfall in computational capacity at The Pennsylvania State University. Although the details of this affliction may not be reproduced elsewhere, it is our opinion that the events we experienced stem from essential, underlying phenomena which do have wide currency. These are, first, that overall demand for computational facilities and services is increasing "exponentially" and shows no sign of slowdown, and, second, that resources (most especially including funds) to provide increases in the relevant supply of computing capacities are not keeping pace and can not be expected to do so.It is possible that technical advances can treat this disorder, but in the nature of the political/bureaucratic systems which are the vehicles for the delivery of such "fixes", acquiring them will not be painless. Concretely, it seems unlikely that faculty and students in colleges and universities can expect relief from recurrent boom-and-bust in computational resources, whether the duration of such cycles is measured in decades or days. It behooves us to ask whether the attendant pains must be endured, and whether they are conducive to easier and more productive use of computing systems. Our answers are, first, that such pain does not ennoble, and, second, that it often is counterproductive. Consequently, we must try to identify the proximate sources of the disrupting effects of these cyclic shortfalls and attempt to curb them, within our means.We propose that the appropriate guidelines for allocating scarce computing resources may be characterized as prescriptions for humane rationing. In the most general terms, these prescriptions are 1) that qualified users should be ensured a fair share of the available resource without unnecessary expenses of effort in competition for them and in queuing, and 2) that use of computing resources should be so governed as to insure that all user sessions are as free as possible of delays, encumbrances, and constraints induced by management practices rather than by inherent limits of hard and software.While rationing is unnecessary during the occasional boom in academic computing resources, we should have on the shelf the management tools which can make fair and effective allocation possible during the recurrent busts we may anticipate in the 1980's.