In many large computer systems with real‐time use (such as the No. 1 Electronic Switching System), the central processing unit handles much of its work through queues. It may spend much of its time cycling through the queues, performing the work requests it finds there. To accomodate varying degrees of urgency, the cycle may visit some hoppers more often than others. (No. 1 ESS strongly relies on this procedure.) This paper provides an approximate method for evaluating different cycles. Using the evaluation method and some approximations, we obtain a formula for the optimum relative frequency with which different queues should be visited. The model used is nonprobabilistic, and treats requests as continuous rather than discrete. The model also ignores certain interdependencies between queues. Despite these drastic simplifications, the results probably provide useful guidance, if interpreted cautiously. © 1969 The Bell System Technical Journal
