Enhanced prefetching and caching strategies for single- and multi-disk systems

Abstract

We study integrated offline caching and prefetching algorithms both in the single- and the multi-disk case. For the problem of minimizing the execution time of a given request sequence, we present simple algorithms. In the single-disk case we present a combinatorial algorithm with an approximation ratio of 3/2. An optimal solution can be computed using a linear program but this requires a very large number of variables. Our new result improves on all previous combinatorial approximation algorithms. For the multi-disk case we give combinatorial 2-approximation algorithms. Additionally, we consider this problem in the resource augmentation model where the approximation algorithms may use more cache lines than the optimal solution they are compared to. Here, we give strategies using one additional cache line per disk that outperform the optimum solution without additional cache in the single-disk case and achieve (1+o(1))-approximations in the multi-disk case. In contrast to some of the previous approaches, all the algorithms we present are combinatorial and easy to implement.