Akram, F., & Zheng, A., & Guo-Yue, A., & Medved, C., & Johnson, C., & Malik, A. W., & Khan, S. U. (2025, March), Performance Unveiled: Comparing Lightweight Devices Testbed and Virtual Machines for Edge Computing  Paper presented at 2025 ASEE Southeast Conference , Mississippi State University, Mississippi. 10.18260/1-2--54186

\bibitem{asee_peer_54186}
  Akram, F., \& Zheng, A., \& Guo-Yue, A., \& Medved, C., \& Johnson, C., \& Malik, A. W., \& Khan, S. U. (2025, March), \emph{Performance Unveiled: Comparing Lightweight Devices Testbed and Virtual Machines for Edge Computing}
  Paper presented at 2025 ASEE Southeast Conference , Mississippi State University, Mississippi.
  10.18260/1-2--54186

Faiza Akram, Andrew Zheng, April Guo-Yue, Cooper Medved, Claire Johnson, Asad Waqar Malik, and Samee U Khan.  "Performance Unveiled: Comparing Lightweight Devices Testbed and Virtual Machines for Edge Computing".  2025 ASEE Southeast Conference , Mississippi State University, Mississippi, 2025, March.  ASEE Conferences, 2025.  https://peer.asee.org/54186 Internet.  16 Jun, 2025

\bibitem{asee_peer_54186}
  Faiza Akram, Andrew Zheng, April Guo-Yue, Cooper Medved, Claire Johnson, Asad Waqar Malik, and Samee U Khan.
  "Performance Unveiled: Comparing Lightweight Devices Testbed and Virtual Machines for Edge Computing".
  \emph{2025 ASEE Southeast Conference , Mississippi State University, Mississippi, 2025, March}.
  ASEE Conferences, 2025.
  https://peer.asee.org/54186 Internet.  16 Jun, 2025

@INPROCEEDINGS{asee_peer_54186
author = "Faiza Akram, Andrew Zheng, April Guo-Yue, Cooper Medved, Claire Johnson, Asad Waqar Malik, and Samee U Khan"
title = "Performance Unveiled: Comparing Lightweight Devices Testbed and Virtual Machines for Edge Computing"
booktitle = "2025 ASEE Southeast Conference "
year = "2025"
month = "March"
address = "Mississippi State University, Mississippi"
publisher = "ASEE Conferences"
note = {https://peer.asee.org/54186}
number = {10.18260/1-2--54186}
}

TY  - CPAPER
AB  - In today’s data-driven world, quick data processing is crucial for IoT applications like e-healthcare, smart cities, and autonomous vehicles, which generate large volumes of real-time data. Efficient processing is essential for improving decision-making and system performance, from detecting health emergencies to managing traffic. As IoT grows, traditional cloud computing struggles to meet real-time demands with its distant centralized servers. Edge computing addresses this by processing data closer to its source, reducing latency, and ensuring quicker, more efficient data handling.
During Summer 2024, as part of the NSF-funded Research Experiences for Undergraduates (REU) program at Mississippi State University, we conducted a study on the role of edge computing in improving data processing efficiency and system resilience using lightweight devices such as Raspberry Pi's. The purpose of this research was to investigate the results in different environments. The study showcases the differences in evaluation results between an application testbed and a virtual environment.
In our experiments, we deployed Apache Storm using Kubernetes and Docker on both a virtual machine and a testbed consisting of Raspberry Pi devices. The virtual machine for the application was hosted on a lab PC with resources identical to the Raspberry Pi hardware to ensure a fair comparison. Both setups processed continuous streams of data from IoT applications, and we measured key performance metrics such as latency and throughput. The experimental results revealed significant differences in performance metrics when deployed at a testbed and a virtual environment, especially in handling high volumes of data generated by IoT applications. 
Interestingly, while edge computing offers the advantage of proximity to data sources, which typically reduces transmission latency, we found that the testbed exhibited higher latency compared to the virtual machine. As per our findings, this can be attributed to several factors, such as virtual machines generally benefit from more efficient resource management, even when hosted on identical hardware. Hypervisors running there can allocate resources dynamically, minimizing delays in processing. Furthermore, they often run with a minimal OS setup, reducing unnecessary overhead, while devices such as Raspberry Pi may face limitations due to their relatively constrained hardware and software. Virtual setups also offer better I/O performance and are optimized for networking, whereas Raspberry Pi's, being single-board computers, may face bottlenecks when dealing with high data throughput. This results in processing data more efficiently in a virtual environment, particularly in environments where high throughput is critical.
Our findings suggest that while Raspberry Pi’s are effective for basic edge computing tasks, more sophisticated hardware is needed to fully leverage the benefits of edge computing in high-performance, real-time applications. Devices with greater processing power, memory, and optimized network capabilities are essential to support the increasing demands of edge computing, particularly as data volumes and throughput requirements grow. These insights focus on the importance of selecting appropriate hardware based on the specific demands of the application environment, with the potential for hybrid approaches combining edge devices with virtualized infrastructure to achieve optimal performance.

AU  - Faiza Akram
AU  - Andrew Zheng
AU  - April Guo-Yue
AU  - Cooper Medved
AU  - Claire Johnson
AU  - Asad Waqar Malik
AU  - Samee U Khan
CY  - Mississippi State University, Mississippi
DA  - 2025/03/09
PB  - ASEE Conferences
TI  - Performance Unveiled: Comparing Lightweight Devices Testbed and Virtual Machines for Edge Computing
UR  - https://peer.asee.org/54186
DO  - 10.18260/1-2--54186
ER  -