Virtualization on Internet of Things Edge Devices With Container Technologies: A Performance Evaluation

TLDR

Lightweight virtualization has transformed software development, and its adoption in IoT and mobile edge computing enables device virtualization on resource‑constrained single‑board computers. This paper seeks to fill the gap by providing a comprehensive performance evaluation of container‑virtualized instances on several low‑power devices for future deployments. The authors evaluate multiple low‑power devices by measuring key performance metrics of containerized workloads to expose strengths and weaknesses. Prior studies on deploying virtualized instances on single‑board computers were limited to few devices and omitted essential performance metrics.

Abstract

Lightweight virtualization technologies have revolutionized the world of software development by introducing flexibility and innovation to this domain. Although the benefits introduced by these emerging solutions have been widely acknowledged in cloud computing, recent advances have led to the spread of such technologies in different contexts. As an example, the Internet of Things (IoT) and mobile edge computing benefit from container virtualization by exploiting the possibility of using these technologies not only in data centers but also on devices, which are characterized by fewer computational resources, such as single-board computers. This has led to a growing trend to more efficiently redesign the critical components of IoT/edge scenarios (e.g., gateways) to enable the concept of device virtualization. The possibility for efficiently deploying virtualized instances on single-board computers has already been addressed in recent studies; however, these studies considered only a limited number of devices and omitted important performance metrics from their empirical assessments. This paper seeks to fill this gap and to provide insights for future deployments through a comprehensive performance evaluation that aims to show the strengths and weaknesses of several low-power devices when handling container-virtualized instances.