Date of Award

3-26-2020

Document Type

Thesis

Degree Name

Master of Science in Cyber Operations

Department

Department of Electrical and Computer Engineering

First Advisor

Mark E. DeYoung, PhD

Abstract

The explosion of Internet Of Things (IoT), embedded and “smart” devices has also seen the addition of “general purpose” single board computers also referred to as “edge devices.” Determining if one of these generic devices meets the need of a new given task however can be challenging. Software generically written to be portable or plug and play may be too bloated to work properly without significant modification due to much tighter hardware resources. Previous work in this area has been focused on micro or chip-level benchmarking which is mainly useful for chip designers or low level system integrators. A higher or macro level method is needed to not only observe the behavior of these devices under a load but ensure they are appropriately configured for the new task, especially as they begin being integrated on platforms with higher cost of failure like self driving cars or drones. In this research we propose a macro level methodology that iteratively benchmarks and optimizes specific workloads on edge devices. With automation provided by Ansible, a multi stage 2k full factorial experiment and robust analysis process ensures the test workload is maximizing the use of available resources before establishing a final benchmark score. By framing the validation tests with a family of network security monitoring applications an end to end scenario fully exercises and validates the developed process. This also provides an additional vector for future research in the realm of network security. The analysis of the results show the developed process met its original design goals and intentions, with the added fact that the latest edge devices like the XAVIER, TX2 and RPi4 can easily perform as an edge network sensor.

AFIT Designator

AFIT-ENG-MS-20-M-062

Share

COinS