Date of Award

12-1992

Document Type

Thesis

Degree Name

Master of Science

Department

Department of Electrical and Computer Engineering

First Advisor

Paul D. Bailor, PhD

Abstract

This research developed a parameterized model that accounts for system overhead and determines when an Ada runtime environment can no longer successfully execute a given Ada task set and still meet all deadlines. The Ada Compiler Evaluation Capability benchmark was used to characterize an actual runtime environment. Using that data, a generic model of a preemptive, rate monotonic priority based runtime system was developed which accounts for overhead due to clock updates, context switching, task suspension, and synchronization. Validation was based on the Hartstone benchmark. First, the benchmark was executed using, the actual runtime environment. Then, those results were compared with the execution of the benchmark using the model. In all cases, except one, the model predicted the point where the task set would fail. A runtime system optimization omitted from model caused the single failure. Experiments conducted using the model allowed the demonstration of the following results. System overhead can be modeled within the existing framework of rate monotonic scheduling theory. Runtime optimizations can be extremely sensitive to phase relationships between task periods and workloads and can render a schedulable task set unschedulable. Requirements of the task set and the performance of the runtime system must be considered simultaneously.

AFIT Designator

AFIT-GCS-ENG-92D-02

DTIC Accession Number

ADA258915

Comments

The author's Vita page is omitted.

Download

Share

COinS