<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN"><!-- saved from url=(0041)http://www.cs.uh.edu/Defenses/format.html -->
University of Houston
Department of Computer Science
In partial fulfillment of the Requirements for the Degree of
Master of Science
Gary A. S. Whittle
will defend his thesis
<span
style="text-transform: uppercase">A HYBRID SCHEME FOR FILE SYSTEM
REFERENCE PREDICTION</span>
Abstract
On-line file prediction schemes have been shown to improve the performance of
file system caches. These methods keep track of previous file access patterns
and use a single heuristic to predict which of the previous successors to the
file being currently accessed is the most likely to be accessed next. We
present here a Hybrid File Prediction
scheme that applies multiple heuristics to this selection problem. We show
that in situations where a single heuristic would be ineffective, at least one
of a set of heuristics provides the sensitivity to make a more accurate
prediction.
We show by analyzing file system traces that keeping even a limited number of successors per file is enough to encompass almost all successors that can be predicted by an optimal on-line predictor. We then analyzed the performance of several proposed successor selection heuristics to determine the relative ability of each to select the correct successor from the successor history.
When compared to the other file prediction schemes tested that rely on a single heuristic for successor selection, the Hybrid File Prediction scheme showed clearly improved performance, especially when the penalty for a missed prediction is increased.
Date: Wednesday,
April 24, 2002
Time: 11:00 AM
Place: 550-PGH
Faculty, students, and the general public are invited.
Thesis Advisor: Dr. Jehan-François Pâris