loic/paper-lowrate-iot
1
0
Fork 0
mirror of https://gitlab.com/manzerbredes/paper-lowrate-iot.git synced 2025-04-08 14:29:07 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

version soumise

Browse source
This commit is contained in:
ORGERIE Anne-Cecile 2019-09-01 16:57:04 +02:00
parent f930372b1f
commit 8ec8558874
2 changed files with 79 additions and 36 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

96
2019-CloudCom.tex
View file

@ -6,6 +6,7 @@
\usepackage{graphicx}
\usepackage{xcolor}
\usepackage{multirow}
\usepackage{amsmath}
\author{\IEEEauthorblockN{Loic Guegan and
Anne-C\'ecile Orgerie}
@ -269,7 +270,7 @@ heating and cooling systems.
\begin{figure}[htbp]
\centering
\includegraphics[width=0.7\linewidth]{./plots/home.png}
\includegraphics[width=0.6\linewidth]{./plots/home.png}
\caption{Overview of IoT devices.}
\label{fig:IoTdev}
\end{figure}
@ -297,7 +298,7 @@ same time.
\begin{figure}[htbp]
\centering
\includegraphics[width=1.\linewidth]{./plots/parts2.png}
\includegraphics[width=.85\linewidth]{./plots/parts2.png}
\caption{Overview of the IoT architecture.}
\label{fig:parts}
\end{figure}
@ -378,7 +379,7 @@ typical round-trip latency of 100ms from the AP to the server~\cite{li_end--end_
Each node from the AP to the Cloud
is a network switch with static and dynamic network energy consumption. The first 8
hops are edge switches and the last one is consider to be a core router as mentioned
in~\cite{jalali_fog_2016}. ECOFEN~\cite{orgerie_ecofen:_2011} is used to model the energy
in~\cite{jalali_fog_2016}. ECOFEN~\cite{orgerie_simulation_2017} is used to model the energy
consumption of the network part. ECOFEN is an ns3 network energy module dedicated to wired
networks. It is based on an energy-per-bit and energy-per-packet
model for the dynamic energy consumption~\cite{sivaraman_profiling_2011,Serrano2015},
@ -404,7 +405,7 @@ power consumption.
\begin{figure}[htbp]
\centering
\includegraphics[width=0.7\linewidth]{./plots/g5k-xp.png}
\includegraphics[width=0.6\linewidth]{./plots/g5k-xp.png}
\caption{Grid'5000 experimental setup.}
\label{fig:g5kExp}
\end{figure}
@ -447,19 +448,33 @@ This is due to the fact that in such a scenario with very small
% Please add the following required packages to your document preamble:
% \usepackage{booktabs}
\begin{table*}[htbp]
%\begin{table*}[htbp]
%\centering
%\caption{Sensors transmission interval effects with 15 sensors}
%\label{tab:sensorsSendIntervalEffects}
%\begin{tabular}{@{}lrrrrr@{}}
%\toprule
%Transmission Interval & 10s & 30s & 50s & 70s & 90s \\ \midrule
%Sensor Power & 13.517\hl{94}W & 13.517\hl{67}W & 13.51767W & 13.51767W & 13.517\hl{61}W \\
%Network Power & 0.441\hl{88}W & 0.441\hl{77}W & 0.44171W & 0.44171W & 0.441\hl{71}W \\
%Application Delay & 0.09951s & 0.10021s & 0.10100s & 0.10203s & 0.10202s \\ \bottomrule
%\end{tabular}
%\end{table*}
\begin{table}[htbp]
\centering
\caption{Sensors transmission interval effects with 15 sensors}
\label{tab:sensorsSendIntervalEffects}
\begin{tabular}{@{}lrrrrr@{}}
\begin{tabular}{@{}lrrr@{}}
\toprule
Transmission Interval & 10s & 30s & 50s & 70s & 90s \\ \midrule
Sensor Power & 13.517\hl{94}W & 13.517\hl{67}W & 13.51767W & 13.51767W & 13.517\hl{61}W \\
Network Power & 0.441\hl{88}W & 0.441\hl{77}W & 0.44171W & 0.44171W & 0.441\hl{71}W \\
Application Delay & 0.09951s & 0.10021s & 0.10100s & 0.10203s & 0.10202s \\ \bottomrule
Transm. Interval & Sensor Power & Network Power & Application Delay \\ \midrule
10s & 13.517\hl{94}W & 0.441\hl{88}W & 0.09951s \\
30s & 13.517\hl{67}W & 0.441\hl{77}W & 0.10021s \\
50s & 13.51767W & 0.44171W & 0.10100s \\
70s & 13.51767W & 0.44171W & 0.10203s \\
90s & 13.517\hl{61}W & 0.441\hl{71}W & 0.10202s \\ \bottomrule
\end{tabular}
\end{table*}
\end{table}
Previous work~\cite{li_end--end_2018} on a similar scenario shows that increasing application
accuracy impacts strongly the energy consumption in the context of data stream analysis. However,
@ -479,7 +494,7 @@ Consequently, sensors energy consumption is dominant, as each sensor adds its ow
\begin{figure}[htbp]
\centering
\includegraphics[width=0.9\linewidth]{./plots/numberSensors-WIFINET.png}
\includegraphics[width=0.65\linewidth]{./plots/numberSensors-WIFINET.png}
\caption{Analysis of the variation of the number of sensors on the IoT/Network part energy consumption for a transmission interval of 10s.}
\label{fig:sensorsNumber}
\end{figure}
@ -498,7 +513,7 @@ the PUE~\cite{Ehsan}.
\begin{figure*}[htbp]
\centering
\includegraphics[width=.7\linewidth]{./plots/vmSize-cloud.png}
\includegraphics[width=.6\linewidth]{./plots/vmSize-cloud.png}
\caption{Server power consumption multiplied by the PUE (= 1.2) using 20 sensors sending data every 10s for various VM memory sizes}
\label{fig:vmSize}
\end{figure*}
@ -514,7 +529,14 @@ consumption. However, the amount of allocated memory to the VM
does not significantly influence the server energy consumption. In
fact, simple database requests do not need any particular
heavy memory accesses and processing time. Thus, remaining experiments are based on VM with 1024MB
of allocated memory.
of allocated memory.
Here, for clarity's sake, we use VMs with one virtual CPU (i.e. one physical CPU core).
The influence of the number of core on the server' energy consumption has been widely
studied in the literature~\cite{heinrich_predicting_2017}. For a given application that scales
smoothly when adding cores, the relation between number of cores and power consumption is
linear. In other words, adding cores for the execution of a parallel application multiplies
accordingly the dynamic energy consumption on the server part.
Next, we study the effects of increasing the number of sensors on the server energy consumption.
Figure~\ref{fig:sensorsNumber-server} presents the results of the average server energy
@ -541,7 +563,7 @@ RAM). This model is detailed in Section~\ref{sec:discuss}.
\begin{figure}[htbp]
\centering
\includegraphics[width=0.7\linewidth]{./plots/sensorsNumberLine-cloud.png}
\includegraphics[width=0.55\linewidth]{./plots/sensorsNumberLine-cloud.png}
\caption{Average server power consumption multiplied by the PUE (= 1.2) for sensors sending data every 10s}
\label{fig:sensorsNumber-server}
\end{figure}
@ -549,7 +571,7 @@ RAM). This model is detailed in Section~\ref{sec:discuss}.
\begin{figure}[htbp]
\centering
\includegraphics[width=0.7\linewidth]{./plots/WPS-cloud.png}
\includegraphics[width=0.55\linewidth]{./plots/WPS-cloud.png}
\caption{Average sensors power cost on the server hosting only our VM with PUE (= 1.2) for sensors sending data every 10s}
\label{fig:sensorsNumber-WPS}
\end{figure}
@ -565,11 +587,15 @@ occur. Therefore, it leads to an increase of the server energy consumption.
\begin{figure*}[htbp]
\centering
\includegraphics[width=0.7\linewidth]{plots/sendInterval-cloud.png}
\includegraphics[width=0.6\linewidth]{plots/sendInterval-cloud.png}
\caption{Server energy consumption multiplied by the PUE (= 1.2) for 50 sensors sending requests at different transmission interval.}
\label{fig:sensorsFrequency}
\end{figure*}
In the next section, we use the hints detailed here and extracted from the
real and simulated experiments in order to provide an end-to-end energy
model that can be used for low-bandwidth IoT applications.
\section{End-to-End Consumption Model}
\label{sec:orgfd3b6ae}
\label{sec:discuss}
@ -597,7 +623,19 @@ consumption. Indeed, in our targeted low-bandwidth IoT application,
the sensor is dedicated to this application. From Table~\ref{tab:params}, one can
derive that the static power
consumption of one IoT sensor is around 0.9 Watts. Its dynamic part
depends on the transmission frequency.
depends on the transmission frequency. So the power consumption of an IoT device:
\begin{footnotesize}
\begin{align*}
P^{IoTdevice} & = P_{static}^{IoT} + P_{dynamic}^{IoT}\\
& = \frac{P_{idle}\times (T - t_{RX}-t_{TX})}{T} + \frac{P_{RX}\times t_{RX} + P_{TX}\times t_{TX}}{T}
\end{align*}
\end{footnotesize}
where \(P_{static}^{IoT}\) and \(P_{dynamic}^{IoT}\) are respectively the static
and dynamic power consumption of the IoT device, $t_{RX}$, $t_{TX}$, and $t_{idle}$ are
the duration spent in each mode (receiving, transmitting and idle) and $P_{RX}$, $P_{TX}$, and $P_{idle}$ the
respective power consumption of each mode, and $T$ is the transmission interval between
two communications from the IoT device to the cloud server.
Concerning the sharing of the network costs, for each router, we
consider its aggregate bandwidth (on all the ports), its average
@ -612,7 +650,7 @@ part as follows:
where \(P_{static}^{device}\) is the static power consumption of the
network device (switch fabrics for instance or gateway),
\(Bandwidth^{application }\) Is the bandwidth used by our IoT application,
\(Bandwidth^{application }\) is the bandwidth used by our IoT application,
\(AggregateBandwidth^{device }\) is the overall aggregated bandwidth of the
network device on all its ports, and \(LinkUtilization^{device}\) is the
effective link utilization percentage. The \(Bandwidth^{application }\)
@ -633,7 +671,6 @@ used in our formula to compute the values that we used in the
simulations and that are presented in left part of Table~\ref{tab:params}.
\begin{table}[htbp]
\centering
\caption{Network Devices Parameters}
@ -653,6 +690,13 @@ simulations and that are presented in left part of Table~\ref{tab:params}.
\end{tabular}
\end{table}
The dynamic consumption of the network part includes a cost
per packet and a cost per byte for each network device as detailed in~\cite{orgerie_simulation_2017}:
\[P_{dynamic}^{netdevice} = \frac{P_{byte}^{device}\times NbBytes + P_{pkt}^{device} \times NbPkts}{T}\]
with $NbBytes$ and $NbPkts$ respectively the number of bytes and packets sent by the application
during one transmission interval and \(P_{byte}^{device}\) and \(P_{pkt}^{device}\) the power consumption
per network device for each byte and each packet respectively.
For the sharing of the Cloud costs, we take into account the number
@ -686,7 +730,7 @@ the previous section.
\begin{figure}[htbp]
\centering
\hspace{1cm}
\includegraphics[width=1.\linewidth]{plots/final.png}
\includegraphics[width=.9\linewidth]{plots/final.png}
\label{fig:end-to-end}
\caption{End-to-end network energy consumption using sensors interval of 10s}
\end{figure}
@ -741,13 +785,16 @@ periodically sending data to a Cloud server using WiFi,
we propose an end-to-end energy consumption model.
This model provides insights on the hidden part of the iceberg: the
impact of IoT devices on the energy consumption of Cloud and network
infrastructures. On our use-case, we show that for a given sensor, its
infrastructures.
On our use-case, we show that for a given sensor, its
larger energy consumption is on the sensor part. But the impact on the
Cloud and network part is huge when using only few sensors with
low-bandwidth applications.
Consequently, with the
IoT exploding growth, it becomes necessary to improve the energy
efficiency of applications hosted on Cloud infrastructures and of IoT devices.
Our future work includes studying other types of IoT wireless
transmission techniques that would be more energy-efficient. We also
plan to study other
@ -755,6 +802,9 @@ IoT applications in order to increase the applicability of our model
and provide advice for increasing the energy-efficiency of IoT infrastructures.
\section*{Acknowledgments}
Experiments presented in this paper were carried out using the Grid'5000 testbed, supported by a scientific interest group hosted by
Inria and including CNRS, RENATER and several Universities as well as other organizations (see \url{https://www.grid5000.fr}).
\bibliographystyle{IEEEtran}
\bibliography{references}

19
references.bib
View file

@ -1284,9 +1284,8 @@ IoT use cases. Index Terms—IoT, Centralized management, Orchestration, ILP, Fo
shorttitle = {Studying the energy consumption of data transfers in {Clouds}},
doi = {10.1109/CloudNet.2014.6968983},
abstract = {Energy consumption is one of the main limiting factors for designing large scale Clouds. Evaluating the energy consumption of Clouds networking architectures and providing multi-level views required by providers and users, is a challenging issue. In this paper, we show how to evaluate and understand network choices (protocols, topologies) in terms of contributions to the energy consumption of the global Cloud infrastructures. By applying the ECOFEN model (Energy Consumption mOdel For End-to-end Networks) and the corresponding simulation framework, we profile and analyze the energy consumption of data transfers in Clouds.},
booktitle = {2014 {IEEE} 3rd {International} {Conference} on {Cloud} {Networking} ({CloudNet})},
booktitle = {{IEEE} {Int.} {Conf.} on {Cloud} {Networking} ({CloudNet})},
author = {Cornea, B. F. and Orgerie, A. C. and Lefèvre, L.},
month = oct,
year = {2014},
keywords = {cloud computing, simulation, power aware computing, Bandwidth, Cloud data transfers, Color, data transfers, ECOFEN model, energy consumption, energy consumption model for end-to-end networks, ethernet networks, global cloud infrastructures, Ports (Computers), Routing protocols, Switches, Transport protocols},
pages = {143--148},
@ -1688,13 +1687,8 @@ IoT use cases. Index Terms—IoT, Centralized management, Orchestration, ILP, Fo
@inproceedings{heinrich_predicting_2017,
title = {Predicting the {Energy}-{Consumption} of {MPI} {Applications} at {Scale} {Using} {Only} a {Single} {Node}},
isbn = {978-1-5386-2326-8},
url = {http://ieeexplore.ieee.org/document/8048921/},
doi = {10.1109/CLUSTER.2017.66},
urldate = {2018-06-15},
publisher = {IEEE},
booktitle = {IEEE Cluster Conference},
author = {Heinrich, Franz Christian and Cornebize, Tom and Degomme, Augustin and Legrand, Arnaud and Carpen-Amarie, Alexandra and Hunold, Sascha and Orgerie, Anne-Cecile and Quinson, Martin},
month = sep,
year = {2017},
pages = {92--102},
file = {Heinrich et al. - 2017 - Predicting the Energy-Consumption of MPI Applicati.pdf:/home/loic/.zotero/zotero/383myqxk.default/zotero/storage/CMDGJDPW/Heinrich et al. - 2017 - Predicting the Energy-Consumption of MPI Applicati.pdf:application/pdf}
@ -2401,9 +2395,8 @@ pages={2818-2823},
language = {en},
number = {5},
urldate = {2019-05-28},
journal = {IEEE Journal on Selected Areas in Communications},
journal = {IEEE J. on Selected Areas in Communications},
author = {Jalali, Fatemeh and Hinton, Kerry and Ayre, Robert and Alpcan, Tansu and Tucker, Rodney S.},
month = may,
year = {2016},
pages = {1728--1739},
file = {Jalali et al. - 2016 - Fog Computing May Help to Save Energy in Cloud Com.pdf:/home/loic/.zotero/zotero/383myqxk.default/zotero/storage/36J4R5W6/Jalali et al. - 2016 - Fog Computing May Help to Save Energy in Cloud Com.pdf:application/pdf}
@ -2448,10 +2441,10 @@ howpublished={\url{https://www.sandvine.com/phenomena}}
@misc{Cisco2019,
author = {Cisco},
title = {{Cisco Visual Networking Index: Forecast and Trends, 2017–2022, White paper}},
title = {{Cisco Visual Networking Index: Forecast and Trends, 2017–2022}},
year = {2019},
month = Feb,
howpublished = {\url{https://www.cisco.com/c/en/us/solutions/collateral/service-provider/visual-networking-index-vni/white-paper-c11-741490.html}}
howpublished = {White paper}
}
@misc{ShiftProject,
@ -2515,7 +2508,7 @@ pages={2822-2846},
@ARTICLE{Zhang2016,
author={Z. {Zhang} and Y. {Bejerano} and S. {Antonakopoulos}},
journal={IEEE/ACM Transactions on Networking},
journal={IEEE/ACM Trans. on Networking},
title={{Energy-Efficient IP Core Network Configuration Under General Traffic Demands}},
year={2016},
volume={24},