Date of Award

1-1-2017

Document Type

Dissertation

Degree Name

Ph.D.

Organizational Unit

Daniel Felix Ritchie School of Engineering and Computer Science, Computer Science

First Advisor

Matthew J. Rutherford, Ph.D.

Second Advisor

Alvaro Arias

Third Advisor

Chris Gauthier-Dickey

Fourth Advisor

Rinku Dewri

Keywords

Cross-platform testing, Embedded systems, Parallel testing, Real-time systems, Unit testing

Abstract

Embedded systems are used in a wide variety of applications (e.g., automotive, agricultural, home security, industrial, medical, military, and aerospace) due to their small size, low-energy consumption, and the ability to control real-time peripheral devices precisely. These systems, however, are different from each other in many aspects: processors, memory size, develop applications/OS, hardware interfaces, and software loading methods. Unit testing is a fundamental part of software development and the lowest level of software testing, as it tests individual or groups of functions, methods, and classes, to increase confidence that the developed software satisfies both software specifications and user requirements. Although hundreds of unit testing frameworks exist, none of them address the diverse properties of real-time embedded platforms. This inspires us to introduce XEUnit, a cross-platform unit testing framework for real-time embedded systems. XEUnit provides scalability to the framework by supporting parallel execution on multiple embedded platforms simultaneously.

To address the time constraints in real-time embedded systems, we evaluate the impact of runtime overhead from traditional instrumentation through a case study of time-sensitive algorithms. Then, we introduce iterative instrumentation, which is a code coverage technique without runtime overhead, along with a case study demonstrating the effectiveness of this technique. Although iterative instrumentation can measure code coverage effectively in time-sensitive applications, the total execution cost of this approach is much higher than traditional instrumentation due to the execution of multiple variants of the system under test. This leads to scalability and performance issues especially in large applications. To solve these issues, there are two approaches we use: reducing the number of variants and executing them simultaneously.

To reduce the number of variants, we present cluster iterative instrumentation, a graph clustering technique that can reduce the number of nodes in a control flow graph resulting in lower execution time. We also provide a case study of node coverage of control software to show the efficiency of cluster iterative instrumentation compared to iterative instrumentation. In addition to reducing the number of variants, the other method is to execute multiple variants at the same time. Because all executions are independent from each other, we can use parallel execution on multiple embedded platforms. Thus, we design and implement a parallel unit testing framework for real-time embedded system along with a case study comparing the execution times from different numbers of embedded platforms (executing nodes).

Our final contribution is a cross-platform unit testing framework using the concepts of runtime adapters and a runtime protocol that enables testers to run code across different embedded platforms. We also demonstrate the effectiveness of this framework by testing black-box test cases on seven different embedded platforms. Overall, our results indicate that cluster iterative instrumentation with parallel unit testing can address the scalability and performance issues, and the case studies demonstrate that XEUnit can effectively test the same code on a variety of embedded platforms.

Publication Statement

Copyright is held by the author. User is responsible for all copyright compliance.

Rights Holder

Tosapon Pankumhang

Provenance

Received from ProQuest

File Format

application/pdf

Language

en

File Size

144 p.

Discipline

Computer science

Download



Share

COinS