Adaptive network resource control

Abstract

An admissions control technique improves processing capacity utilization in a token-based admission control scheme by matching token allocation to actual processing requirements. In an exemplary application, the processing capacity of a processing entity is discretely represented by a plurality of tokens. Work is admitted to the processing entity through a plurality of processing event entry points. Each of these entry points is initially allocated a share of the tokens. During each one of a succession of admission cycles, events are admitted at each entry point until that entry point's share of the tokens is exhausted. At the end of each cycle, tokens are re-allocated to the entry points for use during the next admission cycle based on the actual usage of tokens during the current cycle. An entry point's token allocation may be increased for the next cycle by re-allocating leftover tokens from other, less busy entry points.