University of Houston
Department of Computer Science

In partial fulfillment of the Requirements for the Degree of
Master of Science


Wenjing Chen
will defend her thesis

A Two-Expert Approach for File System Reference Prediction



Abstract

On-line file access prediction schemes are important for improving the performance of file system caches.  We present and evaluate a Two-Expert Approach that combines two separate on-line file access prediction schemes and bases its predictions of the next file to be accessed on the predictions of its two components.  Two variations of the Two-Expert Approach that use the OR operation and the AND operation (the methods of combining the prediction results of the components) are studied and compared.  The performance of the Two-Expert Approach is also examined in comparison with that of its components.  One main conclusion is that the Two-Expert Approach predictor returning a prediction when both components agree on their predictions (the AND operation) outperforms by up to 10 percent its two components whenever prediction accuracy is an important factor and incorrect predictions are to be avoided.  Results for the Two-Expert Approach predictors returning a prediction when either both components agree or one of them does not make a prediction (the OR operation) are less conclusive.

Date: Thursday, October 7, 2004
Time: 5:30 PM
Place: 550-PGH



Faculty, students, and the general public are invited.
Thesis Advisor: Dr. Paris