Date of Award

5-2008

Document Type

Thesis

Degree Name

Master of Science (MS)

Legacy Department

Computer Science

Advisor

Dean, Brian C

Committee Member

Stevenson , D E

Committee Member

Goddard , Wayne

Abstract

Real life servers do not have IID service times - they slow down over time. This is usually dealt with by resetting or replacing the server. Often, however, it is not known how to determine when the best time to reset the server is. Resetting may be a costly process. We make an assumption that only some arrivals (we do not know which) are harmful. Then, we proceed to find an optimal policy for resetting a server with a given service parameterization by observing the service paremeter directly, service times, or waiting times (with and without arrival times) when the harmful arrivals are IID.
Next, we develop some methods for dealing with harmful arrivals that are not IID. Short term memory is used to cope with rapidly changing situations and is found to perform better when each observation contains less information.

Share

COinS