Browse Prior Art Database

File-Based Adaptive Prefetch Caching Design

IP.com Disclosure Number: IPCOM000119628D
Original Publication Date: 1991-Feb-01
Included in the Prior Art Database: 2005-Apr-02
Document File: 2 page(s) / 37K

Publishing Venue

IBM

Related People

Shih, FHW: AUTHOR [+2]

Abstract

Disclosed is a prefetch design for disk caching with dynamic prefetch length selections based on an observed hit/miss history of files during program execution.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 100% of the total text.

File-Based Adaptive Prefetch Caching Design

      Disclosed is a prefetch design for disk caching with
dynamic prefetch length selections based on an observed hit/miss
history of files during program execution.

      When a file is accessed through disk caching, three consecutive
cache hit (1) or miss (0) outcomes are stored in a prefetch status
table.  Using all possible 3-bit histories as transition states, an
eight-state transition diagram is shown in the figure with prefetch
lengths assigned to each transition.  More data blocks, which are
multiples of the cache page size, are prefetched for transitions that
have a tendency to produce additional cache hits.  The prefetch
length for each file access is then determined adaptively according
to the access statistic of that file.

      This design reduces the inefficiency of simple static
prefetching (1) during random disk access.  It also distinguishes the
individual file access pattern from the overall system pattern (2)
and makes a more intelligent dynamic prefetching.

      References:
(1)  A. J. Smith, "Sequential Program Prefetching in Memory
Hierarchies," IEEE Computer 11, 7-21 (December 1978).
(2)  U.S. Patent 4,489,378.