Date of Award

12-1-1996

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Computer Science

First Advisor

Dr. Heshem El-Rewini

Abstract

Real-Time periodic tasks are at the heart of many critical processing computer systems. Nuclear power plants, military command and control systems, aircraft automatic flight control systems, hospital life-support equipment all require precise processing performed within very strict timelines. Missing deadlines can have catastrophic consequences. The rate-monotonic priority assignment policy for scheduling hard-deadline periodic tasks was developed to guarantee those deadlines. In this thesis, we study the problem of scheduling hard-deadline periodic tasks. We begin by surveying the current state of multiprocessor rate-monotonic scheduling and reviewing earlier work. We present the results of a number of experiments we conducted to evaluate the performance of several scheduling heuristics. These heuristics assumed a homogeneous multiprocessing environment. We relax that restriction and introduce three allocation heuristics for scheduling tasks on heterogeneous multiprocessors. Furthermore, we analyze the performance of the proposed algorithms. We compare the quality of the solutions produced by these algorithms and measure them against the optimal solution. Lacking in the current set of real-time multiprocessor heuristics is the awareness of communication between tasks. We add communication into the scheduling model and provide an algorithm to minimize the amount of data transfer between tasks. Furthermore, we examine the performance of this heuristic and compare the schedules it produces with optimal solutions. Lastly, we introduce a scheduling and analysis fool that incorporates several scheduling heuristics. New heuristics are easily added to the tool. The goal of the tool is to help system designers/developers study the performance of different heuristics in scheduling real-time periodic tasks. The tool helps answer “what if” questions, which may also help designers tune their systems to achieve better performance while meeting deadlines.

Comments

A Thesis Presented to the Department of Computer Science and the Faculty of the Graduate College University of Nebraska In Partial Fulfillment of the Requirements for the Degree Master of Science University of Nebraska at Omaha. Copyright 1996 Stephen J. Bernard

Files over 3MB may be slow to open. For best results, right-click and select "save as..."

COinS