IEEE Circuits and Systems Magazine - Q1 2023 - 68

Most energy that is consumed in online video applications is
used by TVs, desktop PCs, and for video encoding.
As another validation step, in future work, one could
set up a complex network of measurement equipment
to measure the power consumption of the various enduser
devices, network nodes, and the DCs performing
online video tasks. Resulting model parameters can
then be used to update the overall energy consumption
estimates reported in this article. The same holds for
the energy consumption of transmission via the global
Internet, where other kinds of data are transmitted at
the same time.
IV. Model Implications
In this section, we will discuss the implications of the
model presented in the previous section and some related
lessons that we can learn. To this end, we take
the point of view of a system engineer whose task it is
to set up an online video platform in the most sustainable,
i.e., energy-efficient manner. At first, in Section
IV-A, we show how the proposed model can help the
system engineer to determine the most important energy
consumers, which strongly depends on the type
of online video service. Subsequently, in Section IV-B,
we discuss a more detailed example for a single video
and show that considering energy consumption when
designing an online video service can lead to significant
energy savings. Finally, in Section IV-C, we unveil future
research directions for improving the energy efficiency
of online video.
Note that the considerations in the next subsections
rely on parameter values that may not be fully
representative of the state-of-the-art hardware used
in practice. As a consequence, the resulting modeling
values may not represent the true energy consumption
of the services. However, from the system engineer's
point of view, the results show clearly which hardware
components and parameters are most important to
Table 5.
Examples of the overall yearly energy consumption of four
different online video services.
System s
On-demand
IPTV
Social Network
Teleconference
68
Overall Es
3.77 TWh
3.83 TWh
4.87 TWh
1.29 TWh
IEEE CIRCUITS AND SYSTEMS MAGAZINE
Es,UT
3.65 TWh
3.65 TWh
94.6 GWh
1.22 TWh
Es,VP
61.7 GWh
111 GWh
4.76 TWh
6.02 GWh
Es,NW
60.2 GWh
65.7 GWh
14.1 GWh
56.9 GWh
address to increase the energy efficiency of an online
video system, which underlines the usefulness of our
proposed model.
A. The Energy Consumption of Online Video
Services
In this section, using the proposed model, we perform
a thought experiment by estimating and comparing the
overall energy consumption of different online video
services. The parameters, i.e., the number of requests
and the corresponding set of devices, are selected
based on the values of an actual on-demand platform
[1]. To allow comparisons with the energy consumption
of other kinds of services, we use similar values for other
services. Hence, the adopted values do not represent
actual services, but help to understand the energy consumption
of the most important components of online
video systems.
Assume that you are a system engineer who is asked
to construct an online video system serving 100 million
end users that employ Ds =⋅
100 106 distinct devices.
Each end user makes use of the service one hour per
day on average. Each video shall be coded with high
visual quality such that the bitrate is set to a value between
br = 2 Mbit
plicity, we adopt a mean value of br = 5 Mbit
s
s
s
lowing. Furthermore, the system engineer can make use
of Σs =1000 servers worldwide to construct and operate
the service using, e.g., a CDN [82].
We are now interested in the expected overall yearly
energy consumption of the online video service. We take
the four typical online video services that are shown
in Table 5 as examples: an on-demand video service, a
social network, an Internet protocol television (IPTV)
service, and teleconferencing. More information on the
choice of the exact values for energy consumption modeling
is provided in detail in Appendix B.
The estimated overall yearly energy
consumption of the four services is listed
in Table 5. We can see that the overall energy
consumption Es
is in the same order
of magnitude for all considered services
(between 1 and 4 TWh) and corresponds
to roughly one percent of the total electrical
power production of Germany in
the year 2019 [83], which corresponds to
GHG emissions of approximately 2.5 million
tons CO2E [84]. Please note that these
FIRST QUARTER 2023
and br =10 Mbit
[16], where for simin
the fol
IEEE Circuits and Systems Magazine - Q1 2023

Table of Contents for the Digital Edition of IEEE Circuits and Systems Magazine - Q1 2023
