Degree Type

Dissertation

Date of Award

1983

Degree Name

Doctor of Philosophy

Department

Electrical and Computer Engineering

Abstract

A technique to support the operating system has been presented. The approach utilizes a modular microprocessor based support system to parallel process and/or assist some operating system funtional units. Adopting the proposed technique may result in reducing the amount of processing power devoted to the execution of the operating system code as well as in reducing overall system cost;Two application examples of the proposed technique have been invented. The first is a support module for Habermann's deadlock avoidance algorithm. Without the support module, the algorithm would have an execution time O(m('2)), where m is the number of processes in the system. The design and operation of a support module that employs m + 1 microprocessors and is capable of performing the algorithm with an execution time O(m) has been described. Another module utilizes (m/k) + 1 microprocessors, and has an execution time O(km), k > 1, has been discussed. This latter module is suitable when m is very large. A fast module that has an execution time O(km), k < 1, and is appropriate for systems with small m has also been discussed;The second application example is a support module for the exact implementation of the Least Recently Used (LRU) replacement policy in a demand paging memory management system. It was believed that the exact implementation of the LRU was "not feasible" because it would represent a tremendous overhead. This is no longer true if the proposed technique is considered. A submodule that employs an MC68000 microprocessor was designed and built to execute the exact LRU function. To test the submodule, it was found necessary to build an address stream generator module to simulate the main system. Therefore, an address stream generation module was built around another MC68000 microprocessor. The LRU submodule was then extensively tested under different address arrival rates. The data obtained from the experiments prove the feasibility of implementing the exact LRU algorithm at a very low cost. It also largely endorses the proposed support technique.

DOI

https://doi.org/10.31274/rtd-180813-6050

Publisher

Digital Repository @ Iowa State University, http://lib.dr.iastate.edu/

Copyright Owner

Ahmed Amin Elamawy

Language

en

Proquest ID

AAI8323279

File Format

application/pdf

File Size

137 pages

Share

COinS