On paging with locality of reference

Abstract

Motivated by the fact that competitive analysis yields too pessimistic results when applied to the paging problem, there has been considerable research interest in refining competitive analysis and in developing alternative models for studying online paging.In this paper, we propose a new, simple model for studying paging with locality of reference. The model is closely related to Denning's working set concept and directly reflects the amount of locality that request sequences exhibit. We use the page fault rate to evaluate the quality of paging algorithms, which is the performance measure used in practice.We develop tight or nearly tight bounds on the fault rates achieved by popular paging algorithms such as LRU, FIFO, deterministic Marking strategies and LFD. These bounds show that LRU is an optimal online algorithm, whereas FIFO and Marking strategies are not optimal in general. We present an experimental study comparing the page fault rates proven in our analyses to the page fault rates observed in practice.

List Update with Locality of Reference

“…As stated in the Introduction, several models have been proposed for paging with locality of reference [23,2,8]. In this paper, we consider the model of Albers et al [2], in which a request sequence has high locality of reference if the number of distinct requests in a window of size n is small.…”

“…In this paper, we consider the model of Albers et al [2], in which a request sequence has high locality of reference if the number of distinct requests in a window of size n is small. In Section 4 we will present experimental evidence which supports the validity of this model for the list update problem.…”

“…Consider a function that represents the maximum number of distinct items in a window of size n, on a given request sequence. For the paging problem, extensive experiments with real data show that this function can be bounded by a concave function for most practical request sequences [2]. Let f be an increasing concave function.…”

“…Informally, this property suggests that the currently requested item is likely to be requested again in the near future. For the paging problem, several models for capturing locality of reference have been proposed [23,2,8]. Input sequences of list update algorithms in practice show locality of reference [15,21,10] and on-line list update algorithms try to take advantage of this property [15,20].…”

“…We begin by showing that all on-line list update algorithms are equivalent according to Bijective Analysis under the modified cost model. We then extend a model for locality of reference, proposed by Albers et al [2] in the context of the paging problem to the list update problem. The validity of the extended model is supported by experimental results obtained on the Calgary Corpus, which is frequently used as a standard benchmark for evaluating the performance of compression algorithms (and by extension list update algorithms, e.g.…”