Date of Award

Summer 1998

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Computing Sciences - (Ph.D.)

Department

Computer and Information Science

First Advisor

Alexander D. Stoyenko

Second Advisor

Murat Tanik

Third Advisor

Sanjoy Kumar Baruah

Fourth Advisor

Michael G. Hinchey

Fifth Advisor

Phillip A. Laplante

Sixth Advisor

Peter A. Ng

Seventh Advisor

Donald H. Sebastian

Abstract

While the management of resources in computer systems can greatly impact the usefulness and integrity of the system, finding an optimal solution to the management problem is unfortunately NP hard. Adding to the complexity, today's 'modern' systems - such as in multimedia, medical, and military systems - may be, and often are, comprised of interacting real and non-real-time components. In addition, these systems can be driven by a host of non-functional objectives – often differing not only in nature, importance, and form, but also in dimensional units and range, and themselves interacting in complex ways. We refer to systems exhibiting such characteristics as Complex Systems (CS).

We present a method for handling the multiple non-functional system objectives in CS, by addressing decomposition, quantification, and evaluation issues. Our method will result in better allocations, improve objective satisfaction, improve the overall performance of the system, and reduce cost -in a global sense. Moreover, we consider the problem of formulating the cost of an allocation driven by system objectives. We start by discussing issues and relationships among global objectives, their decomposition, and cost functions for evaluation of system objective. Then, as an example of objective and cost function development, we introduce the concept of deadline balancing. Next, we proceed by proving the existence of combining models and their underlying conditions. Then, we describe a hierarchical model for system objective function synthesis. This synthesis is performed solely for the purpose of measuring the level of objective satisfaction in a proposed hardware to software allocation, not for design of individual software modules. Then, Examples are given to show how the model applies to actual multi-objective problems.

In addition the concept of deadline balancing is extended to a new scheduling concept, namely Inter-Completion-Time Scheduling (ICTS. Finally, experiments based on simulation have been conducted to capture various properties of the synthesis approach as well as ICTS. A prototype implementation of the cost functions synthesis and evaluation environment is described, highlighting the applicability and usefulness of the synthesis in realistic applications.

Share

COinS