Đang chuẩn bị liên kết để tải về tài liệu:
THE FRACTAL STRUCTURE OF DATA REFERENCE- P4

THE FRACTAL STRUCTURE OF DATA REFERENCE- P4:For purposes of understanding its performance, a computer system is traditionally viewed as a processor coupled to one or more disk storage devices, and driven by externally generated requests (typically called transactions). Over the past several decades, very powerful techniques have become available to the performance analyst attempting to understand, at a high level, the operational behavior of such systems. | Chapter 1 HIERARCHICAL REUSE MODEL The ability ofcomputers to rapidly deliver requested information at the time it is needed depends critically on the memory hierarchy through which data stored on disk is made available to the processor. The organization of storage into a hierarchy has made possible today s spectacular pace of technology evolution in both the volume of data and the average speed of data access. But the effectiveness of a memory hierarchy depends in turn upon the patterns of data reference that it must support. The purpose of this chapter is to develop a practical description of data reference patterns suitable for use in analyzing the performance of a memory hierarchy. After filling in some essential background and terminology we shall first show that the traditional memoryless model of interarrivals does not offer the practical description we are looking for. Indeed this approach is not even able to predict something as fundamental as the advantages offered by organizing memory into a hierarchy which every computer architect today takes for granted on the basis ofpractical experience. As an alternative we develop the hierarchical reuse model. This description ofdata reference uses a heavy-tailed distribution of interarrival times to reflect transient patterns of access. We show that such an approach both makes sense and fits well with empirical data. Its power as a tool for analysis becomes apparent as we study the cache visits made by data stored on disk and the resulting requirements for cache memory. Finally we extend our analysis to a memory hierarchy containing three or more levels with emphasis on the important special case in which disk storage requests are buffered using both host file buffers as well as storage control cache. 2 THE FRACTAL STRUCTURE OF DATA REFERENCE 1. BACKGROUND Although the concepts discussed in the present chapter apply broadly to memory and storage hierarchies within many computing platforms we shall draw our .