Date of Award

Fall 1996

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

Sanjoy Kumar Baruah

Third Advisor

Tadao Ichikawa

Fourth Advisor

Phillip A. Laplante

Fifth Advisor

C. L. Liu

Sixth Advisor

Thomas J. Marlowe

Seventh Advisor

James A. McHugh

Eighth Advisor

Peter A. Ng

Ninth Advisor

C. L. Liu

Abstract

Both time constraints and logical correctness are essential to real-time systems and failure to specify and observe a time constraint may result in disaster. Two orthogonal issues arise in the design and analysis of real-time systems: one is the specification of the system, and the semantic model describing the properties of real-time programs; the other is the scheduling and allocation of resources that may be shared by real-time program modules.

The problem of scheduling tasks with precedence and timing constraints onto a set of processors in a way that minimizes maximum tardiness is here considered. A new scheduling heuristic, Least Space Time First (LSTF), is proposed for this NP-Complete problem. Basic properties of LSTF are explored; for example, it is shown that (1) LSTF dominates Earliest-Deadline-First (EDF) for scheduling a set of tasks on a single processor (i.e., if a set of tasks are schedulable under EDF, they are also schedulable under LSTF); and (2) LSTF is more effective than EDF for scheduling a set of independent simple tasks on multiple processors.

Within an idealized framework, theoretical bounds on maximum tardiness for scheduling algorithms in general, and tighter bounds for LSTF in particular, are proven for worst case behavior. Furthermore, simulation benchmarks are developed, comparing the performance of LSTF with other scheduling disciplines for average case behavior.

Several techniques are introduced to integrate overhead (for example, scheduler and context switch) and more realistic assumptions (such as inter-processor communication cost) in various execution models. A workload generator and symbolic simulator have been implemented for comparing the performance of LSTF (and a variant -- LSTF+) with that of several standard scheduling algorithms.

LSTF's execution model, basic theories, and overhead considerations have been defined and developed. Based upon the evidence, it is proposed that LSTF is a good and practical scheduling algorithm for building predictable, analyzable, and reliable complex real-time systems.

There remain some open issues to be explored, such as relaxing some current restrictions, discovering more properties and theorems of LSTF under different models, etc. We strongly believe that LSTF can be a practical scheduling algorithm in the near future.

Share

COinS