Sleepy Network Devices: Implementation and Energy Modeling of Low-Power Proxying Solutions
(2024) EITM01 20241Department of Electrical and Information Technology
- Abstract
- As the industrial, commercial, and domestic prevalence of Internet of Things networks grows, so does the search for viable solutions to decrease their energy consumption. One common solution is using a low-power proxy that can control the power state of the devices within the network. To evaluate such solutions, either implementations can be made and measured for energy consumption, or energy modeling may be used to estimate potential solutions. To test both of these methods, previously proposed low-power proxying approaches were prototyped, namely single-, multi-, and parallel-proxies. These proxies were tested within networks of Power-over-Ethernet surveillance cameras in different scenarios. To validate the accuracy of energy modeling,... (More)
- As the industrial, commercial, and domestic prevalence of Internet of Things networks grows, so does the search for viable solutions to decrease their energy consumption. One common solution is using a low-power proxy that can control the power state of the devices within the network. To evaluate such solutions, either implementations can be made and measured for energy consumption, or energy modeling may be used to estimate potential solutions. To test both of these methods, previously proposed low-power proxying approaches were prototyped, namely single-, multi-, and parallel-proxies. These proxies were tested within networks of Power-over-Ethernet surveillance cameras in different scenarios. To validate the accuracy of energy modeling, estimations of the energy consumption and efficiency were made and compared against the measured results. The parallel proxy approach emerged as the only energy-efficient method from the tests conducted, reducing consumption by up to 26.32 \%. In contrast, single- and multi-proxy methods led to higher power consumption, due to issues related to peripheral power sourcing equipment. Model estimates showed significant improvement after error correction. The energy model had an RMSE of 82.52 watt-hours when estimating energy consumption, which after error correction dropped to 30.63 watt-hours. Estimating relative change in energy consumption initially showed a RMSE of 9.37 percentage points, which increased to 10.39 percentage points after error correction. All-though $R^2$ seemed to point towards a good fit. (Less)
- Popular Abstract
- Putting Network Devices to Sleep
Lowering the electrical usage of networks by inactivating cameras while idle, and automatically awakening them when they are needed using low-power proxies.
Internet of Things (IoT) networks are becoming increasingly more prevalent in every sector, from domestic households to corporate and governmental settings. They are used in many different environments and scenarios and fulfill different purposes. One example is the use of network surveillance cameras to do everything from recording road traffic and accidents to surveillance of buildings during closing hours. Unfortunately, the cameras and devices used within the network consume a substantial amount of energy if left powered on constantly. An... (More) - Putting Network Devices to Sleep
Lowering the electrical usage of networks by inactivating cameras while idle, and automatically awakening them when they are needed using low-power proxies.
Internet of Things (IoT) networks are becoming increasingly more prevalent in every sector, from domestic households to corporate and governmental settings. They are used in many different environments and scenarios and fulfill different purposes. One example is the use of network surveillance cameras to do everything from recording road traffic and accidents to surveillance of buildings during closing hours. Unfortunately, the cameras and devices used within the network consume a substantial amount of energy if left powered on constantly. An approach implemented in this project aiming to decrease this waste of electricity showed promising energy efficiency in several tests. Some measurements were in the range of 25 \% to 30 \% decrease in energy consumption by simply installing a single device within a network. Energy models used in estimating consumption and efficiency of such approaches were created and showed some accuracy with an average error as low as 5.40 percentage points. Such models are promising tools to be used in assisting evaluation of future potential solutions before initial implementation stages.
In the IoT networks of today, a requirement of devices on these networks is usually that they should be available for service at any moment, also known as having network presence. If a device is unavailable, due to being powered off or disconnected, work cannot be performed, and the device may be viewed as inaccessible meaning it will not have work assigned to it in the future. Within networks of devices where there might be long intervals of time spent in between such instances of work, this means the devices must idly stand by while continuously consuming electricity to uphold their presence within the network.
Imagine if there was a way to ensure that devices, such as surveillance cameras, were only powered on when they were needed. One way would be to have one person at each camera, turning them on when there is activity in the surveillance area. Obviously, this would not be feasible, especially in larger networks. Instead, a proposed method is using a so-called low-power proxy device that can do this job. This device could act as the intermediary between the cameras and the rest of the network, ensuring that the camera is active only when needed and otherwise either powered off or suspended. Several different low-power proxies, the single-, multi-, and parallel-proxy, each with different benefits and drawbacks, were implemented with some showing the promising energy efficiency as presented before. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9177415
- author
- Cederberg, Oscar LU and Ahrendtsen Blom, Kristian LU
- supervisor
- organization
- course
- EITM01 20241
- year
- 2024
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- low-power proxy, green proxy, energy consumption, power consumption, energy consumption modelling, modelling, residual plot, model accuracy, Power over Ethernet, Wake-on-LAN, network presence, Internet of Things, embedded systems, surveillance cameras
- report number
- LU/LTH-EIT 2024-1028
- language
- English
- id
- 9177415
- date added to LUP
- 2024-11-21 15:33:30
- date last changed
- 2024-11-21 15:33:30
@misc{9177415,
abstract = {{As the industrial, commercial, and domestic prevalence of Internet of Things networks grows, so does the search for viable solutions to decrease their energy consumption. One common solution is using a low-power proxy that can control the power state of the devices within the network. To evaluate such solutions, either implementations can be made and measured for energy consumption, or energy modeling may be used to estimate potential solutions. To test both of these methods, previously proposed low-power proxying approaches were prototyped, namely single-, multi-, and parallel-proxies. These proxies were tested within networks of Power-over-Ethernet surveillance cameras in different scenarios. To validate the accuracy of energy modeling, estimations of the energy consumption and efficiency were made and compared against the measured results. The parallel proxy approach emerged as the only energy-efficient method from the tests conducted, reducing consumption by up to 26.32 \%. In contrast, single- and multi-proxy methods led to higher power consumption, due to issues related to peripheral power sourcing equipment. Model estimates showed significant improvement after error correction. The energy model had an RMSE of 82.52 watt-hours when estimating energy consumption, which after error correction dropped to 30.63 watt-hours. Estimating relative change in energy consumption initially showed a RMSE of 9.37 percentage points, which increased to 10.39 percentage points after error correction. All-though $R^2$ seemed to point towards a good fit.}},
author = {{Cederberg, Oscar and Ahrendtsen Blom, Kristian}},
language = {{eng}},
note = {{Student Paper}},
title = {{Sleepy Network Devices: Implementation and Energy Modeling of Low-Power Proxying Solutions}},
year = {{2024}},
}