摘要: A program's working set W(t, T) at time t is the of distinct pages among T most recently referenced pages. Relations between average working-set size, missing-page rate, and interreference-interval distribution may be derived both from time-average definitions ensemble-average (statistical) definitions. An efficient algorithm for estimating these quantities given. The relation to LRU (lease used) paging characterized. independent-reference model, in which page references are statistically independent, used assess effects interpage dependencies on size observations. Under general assumptions, shown normally distributed.