Sustainability vs. Quality of Experience: Striking the Right Balance for Video Streaming

The exponential growth in internet data traffic, driven by the widespread use of video streaming applications, has resulted in increased energy consumption and carbon emissions. This outcome is primarily due to higher resolution or higher framerates content and the ability to watch videos on various end-devices. However, efforts to reduce energy consumption in video streaming services may have unintended consequences on users’ Quality of Experience (QoE). This column delves into the intricate relationship between QoE and energy consumption, considering the impact of different bit rates on end-devices. We also consider other factors to provide a more comprehensive understanding of whether these end-devices have a significant environmental impact. It is essential to carefully weigh the trade-offs between QoE and energy consumption to make informed decisions and develop sustainable practices in video streaming services.

Energy Consumption for Video Streaming

In the past few years, we have seen a remarkable expansion in how online content is delivered. According to Sandvine’s 2023 Global Internet Phenomena Report [1], video usage on the Internet has increased by 24% in 2022 and now accounts for 65% of all Internet traffic. This surge in video usage is mainly due to the growing popularity of streaming video services. Videos have become an increasingly popular form of online content, capturing a significant portion of internet users’ attention and shaping how we consume information and entertainment online. Therefore, the rising quality expectations of end-users have necessitated research and implementation of video streaming management approaches that consider the Quality of Experience (QoE) [2]. The idea is to develop applications that can work within the energy and resource limits of end-devices, while still delivering the Quality of Service (QoS) needed for smooth video viewing and an improved user experience (QoE). Even though video streaming services are advancing so quickly, energy consumption is still a significant issue causing many concerns about its impact and the urgent need to boost energy efficiency [14].

The literature provides four main elements: the data centres, the data transmission networks, the end-devices and the consumer behaviour analysing of the energy consumption of video streaming [3]. In this regard, in [4], the authors present a comprehensive review of existing literature on the energy consumption of online video streaming services. Then, they outline the potential actions that can be taken by both service providers and consumers to promote sustainable video streaming, drawing from the literature studies discussed. Their summary of the current possible actions for sustainable video streaming, from both the provider’s and consumer’s perspective, is expressed in the following segments with some of the possible solutions:

• Data center: CDN (Content Delivery Network) can be utilized to offload contents/applications to the edge from the provider’s side and choose providers that prioritize sustainability from the consumer’s side.
• Data transmission network: Data compression/encoding algorithms from the provider’s side and no autoplay from the consumer’s side.
• End-Device: Produce energy-efficient devices from the provider’s size and prefer small-size (mobile) devices from the consumer’s side.
• Consumer behaviour: Reduce the number of subscribers from the provider’s size and prefer watching videos with other people than alone from the consumer’s side.

Finally, they noted that the end device and consumer behaviour are the primary contributors to energy costs in the video streaming process. This result includes actions such as reducing video resolution and using smaller devices. However, taking such actions may have a potential downside as they can negatively impact the QoE due to their effect on video quality. Therefore, in [5], they found that by sacrificing the maximum QoE and aiming for good quality instead (e.g., MOS score of 4=Good instead of MOS score 5=Excellent), significant energy savings can be achieved in video-conferencing services. This is possible by using lower video bitrates compared to higher bitrates which result in higher energy consumption, as per their logarithmic QoE model. Regarding this research, in [4], the authors propose identifying an acceptable level of QoE, rather than striving for maximum QoE, as a potential solution to reduce energy consumption while still meeting consumer satisfaction. They conducted a crowdsourcing survey to gather real consumer opinions on their willingness to save energy consumption while streaming online videos. Then, they analysed the survey results to understand how willing people are to lower video streaming quality in order to achieve energy savings.

Green Video Streaming: The Trade-Off Between QoE and Energy Consumption

To provide a trade-off between QoE and Energy Consumption, we looked at the connection between video bitrate of standard resolution, electricity usage, and perceived QoE for a video streaming service on four different devices (smartphone, tablet, laptop/PC, and smart TV) as taken from [4].

They calculated the energy consumption of streaming on devices which is provided in [6]: Q_i = t_i.(P_i+R_i.ƿ), in the given equation, Q_i represents the electricity consumption (in kWh) of the i-th device, t_i denotes the streaming duration (in hours per week) for the i-th device, P_i represents the power load (in kW) of the i-th device, R_i signifies the data traffic (in GB/h) for a specific bitrate, and ρ = 0.1 kWh/GB represents the electricity intensity of data traffic.

Then,  to estimate the perceived QoE based on the video bitrate, the authors employed a QoE model from [7], as noted in their analysis which is: QoE = a.br^b + c, where “br” represents the bitrate, and “a”, “b”, and “c” are the regression coefficients calculated for specific resolutions.

After taking this into account, we can establish a link between the QoE model, energy consumption, and the perceived QoE associated with video bitrate across various end-devices. Therefore, we implemented the green QoE model in [8] to provide a trade-off between the perceived QoE and the calculated energy consumption from above in the following way: f_γ(x)= 4/(log(x’_5)-log(x_1))*log(x)+ (log(x’_5)-5*log(x_1))/(log(x’_5)-log(x_1)). The given equation represents the mapping function between video bitrate and Mean Opinion Scores (MOS), considering both the minimum bitrate x_1 corresponding to MOS 1 and the maximum bitrate x_5 corresponding to MOS 5. Moreover, the factor γ, representing the greenness of a user, is considered in the context of maximum bitrate x’_5 = x_5/γ, which results in a MOS score of 5.

The model focuses on the concept of a “green user,” who considers the energy consumption aspect in their overall QoE evaluations. Thus, a green user might rate their QoE slightly lower in order to reduce their carbon footprint compared to a high-quality (HQ) user (or “non-green” user) who prioritizes QoE without considering energy consumption.

The numerical results for the energy consumption (in kWh) and the MOS scores depending on the video bitrate can be simplified with linear and logarithmic regressions, respectively. In Figure 1, the graph depicts a linear regression analysis conducted to examine the relationship between energy consumption (kWh) and bitrate (kbps). The y-axis represents energy consumption while the x-axis represents bitrate (kbps). The graph displays a straight-line trend that starts at 1.6 kWh and extends up to 3.5 kWh as the bitrate increases. The linear fitting function used for the analysis is formulated as: kWh = f(bitrate) = a * bitrate + c, where ‘a’ represents the slope and ‘c’ represents the y-intercept of the line.

Figure 1 visually illustrates how energy consumption tends to increase with higher bitrates, as indicated by the positive slope of the linear regression line in Figure 1. One notable observation is that as video bitrates increase, the electricity consumption of end-devices also tends to increase. This can be attributed to the larger amount of data traffic generated by higher-resolution video content, which requires higher bitrates for transmission. Consequently, smart TVs are likely to consume more energy compared to other devices. This finding is consistent with the results obtained from the linear regression model, as described in [4], further validating the relationship between bitrate and energy consumption.

As illustrated in Figure 2, the relationship between MOS and video bitrate (kbps) follows a logarithmic pattern. Therefore, we can use a straightforward QoE model to estimate the MOS if there is information about the video bitrate. This can be achieved by utilizing a logistic regression model MOS(x), where MOS = f(x) = a * log(x) + c, with x representing the video bitrate in Mbps, and a and c being coefficients, as provided in [9]. After, MOS and video bitrate (kbps) values in [4] are applied to the above-mentioned QoE green model equation regarding the logistic regression model, which is an extension of the logarithmic regression model [8]. This relationship allows to determine the green user QoE model and we exemplary show the green user QoE model for smart TV (using γ=2 in f_γ(x)).

According to Figure 2, it is categorized users into two groups: those who prioritize high-quality (HQ) video regardless of energy consumption, and green users who prioritize energy efficiency while still being satisfied with slightly lower video quality. It can be observed that the MOS value changes in video quality on their smart TVs faster compared to other end-devices.  This is evident from the steeper curve in the smart TV section. On the other hand, when looking at the curve for tablets, it shows that changes in bitrate have a milder impact on MOS values. The outcome suggests that video streaming on smaller screens, such as tablets or laptops, may contribute less to the perception of quality changes. Considering that those small-screen devices consume less energy than larger screen devices, it may be preferable to use lower resolution videos instead of high-resolution ones. Analysing the relationship between laptops and tablets, it can be seen that low-resolution video streaming on laptops resulted in lower MOS scores compared to the tablet. From this result, it can be inferred that the choice of end-device and user behaviour plays a significant role in energy savings. Figure 2 indicates that the MOS values for the green user of a smart TV is comparable to the MOS values of an HQ user using a laptop.

Concerning this outcome, in [10], the authors presented the results of a subjective assessment aimed at investigating how different factors, such as video resolution, luminance, and end devices (TV, Laptop, and Smartphone), impact the QoE and energy consumption of video streaming services. The study found that, in certain conditions such as dark or bright environments, low device backlight luminance, or small-screen devices), users may need to strike a balance between acceptable QoE and sustainable (green) choices, as consuming more energy (e.g., by streaming higher-quality videos) may not significantly enhance the QoE.

Therefore, Figure 3 plots the trade-off relationship between energy consumption (kWh) and MOS for the end devices (such as smart TV, laptop and tablet). Thereby, we differentiate the HQ user and the green user, which presents some interesting results. First, a MOS score of 4 leads to comparable energy consumption results for green and HQ users. The relative differences are rather small. However, aiming for best quality (MOS 5) leads to significant differences. Furthermore, it is seen that the device type has a significant impact on energy consumption. Even for green users, which rate lower bitrates with higher MOS scores than HQ users, the energy consumption of the smart TV is much higher than for any quality (i.e. bitrate) for laptop and tablet users. Thus, device type and user behaviour are essential to strike the right balance between QoE and energy consumption.

Discussions and Future Research

Meeting the QoE expectations of end-users is essential to fulfilling the requirements of video streaming services. As users are the primary viewers of streaming videos in most real-world scenarios, subjective QoE assessment [11] provides a direct and dependable means to evaluate the perceptual quality of video streaming. Furthermore, there is a growing need to create objective QoE assessment models provided in [12]– [13]. However, many studies have focused on investigating the QoE obtained through subjective and objective models and have overlooked the consideration of energy consumption in video streaming.

Therefore, in the previous section, we have discussed how the different elements within the video streaming ecosystem play a role in consuming energy and emitting CO2.  The findings pave the way for an objective response to determining an appropriate optimal video bitrate for viewing, considering both QoE and sustainability considerations, which can be further explored in future research.

It is evident that addressing energy consumption and emissions is crucial for the future of video streaming systems, while ensuring that end-users’ QoE is not compromised poses a significant and ongoing challenge. Thus, potential solutions to prevent energy consumption increase in QoE while still satisfying the user include streaming videos on smaller screen devices and watching lower resolution videos that offer sufficient quality instead of the highest resolution ones. Here, it can be highlighted the importance of user behavior to prevent energy consumption. Additionally, trade-off models can be developed using the green QoE model (especially for smarTV) by identifying ideal bitrate values for energy savings and user satisfaction in the QoE.

Delving deeper into the dynamics of the video streaming ecosystem, it becomes increasingly clear that energy consumption and emissions are critical concerns that must be addressed for the sustainable future of video streaming systems. The environmental impact of video streaming, particularly in terms of carbon emissions, cannot be understated. With the growing awareness of the urgent need to combat climate change, mitigating the environmental footprint of video streaming has become a pressing priority.

As video streaming technologies evolve, optimizing energy-efficient approaches without compromising users’ QoE is a complex task. End-users, who expect seamless and high-quality video streaming experiences, should not be deprived of their QoE while addressing the energy and emissions concerns. The outcome opens a novel door for an objective answer to the question of what constitutes an appropriate optimal video bitrate for viewing that takes into account both QoE and sustainability concerns.

Future research in this area is crucial to explore innovative techniques and strategies that can effectively reduce the energy consumption and carbon emissions of video streaming systems without sacrificing the QoE. Additionally, collaborative efforts among stakeholders, including researchers, industry practitioners, policymakers, and end-users, are essential in devising sustainable video streaming solutions that consider both environmental and user experience factors [14].

In conclusion, the discussions on the relationship between energy consumption, emissions, and QoE in video streaming systems emphasize the need for continued research and innovation to achieve a sustainable balance between environmental sustainability and user satisfaction.

References

• [1] Sandvine. The Global Internet Phenomena Report. January 2023. Retrieved April 24, 2023
• [2] M. Seufert, S. Egger, M. Slanina, T. Zinner, T. Hoßfeld and P. Tran-Gia, “A Survey on Quality of Experience of HTTP Adaptive Streaming,” in IEEE Communications Surveys & Tutorials, vol. 17, no. 1, pp. 469-492, Firstquarter 2015, doi: 10.1109/COMST.2014.2360940., 2015.
• [3] Reinhard Madlener, Siamak Sheykhha, Wolfgang Briglauer,”The electricity- and CO2-saving potentials offered by regulation of European video-streaming services,” Energy Policy,vol. 161, p. 112716, 2022.
• [4] G. Bingöl, S. Porcu, A. Floris and L. Atzori, “An Analysis of the Trade-off between Sustainability,” in IEEE ICC Workshop-GreenNet, Rome, 2023.
• [5] T. Hoßfeld, M. Varela, L. Skorin-Kapov, P. E. Heegaard, “What is the trade-off between CO2 emission and video-conferencing QoE?,” ACM SIGMM Records, 2022.
• [6] P. Suski, J. Pohl, and V. Frick, “All you can stream: Investigating the role of user behavior for greenhouse gas intensity of video streaming,” in Proc. of the 7th Int. Conf. on ICT for Sustainability, 2020, pp. 128–138.
• [7] M. Mu, M. Broadbent, A. Farshad, N. Hart, D. Hutchison, Q. Ni, and N. Race, “A Scalable User Fairness Model for Adaptive Video Streaming Over SDN-Assisted Future Networks,” IEEE Journal on Selected Areas in Communications, vol. 34, no. 8, p. 2168–2184, 2016.
• [8] T. Hossfeld, M. Varela, L. Skorin-Kapov and P. E. Heegaard, “A Greener Experience: Trade-offs between QoE and CO2 Emissions in Today’s and 6G Networks,” IEEE Communications Magazine, pp. 1-7, 2023.
• [9] J. P. López, D. Martín, D. Jiménez and J. M. Menéndez, “Prediction and Modeling for No-Reference Video Quality Assessment Based on Machine Learning,” in 14th International Conference on Signal-Image Technology & Internet-Based Systems (SITIS), IEEE, pp. 56-63, Las Palmas de Gran Canaria, Spain, 2018.
• [10] G. Bingöl, A. Floris, S. Porcu, C. Timmerer and L. Atzori, “Are Quality and Sustainability Reconcilable? A Subjective Study on Video QoE, Luminance and Resolution,” in 15th International Conference on Quality of Multimedia Experience (QoMEX), Gent, Belgium, 2023.
• [11] G. Bingol, L. Serreli, S. Porcu, A. Floris, L. Atzori, “The Impact of Network Impairments on the QoE of WebRTC applications: A Subjective study,” in 14th International Conference on Quality of Multimedia Experience (QoMEX), Lippstadt, Germany, 2022.
• [12] D. Z. Rodríguez, R. L. Rosa, E. C. Alfaia, J. I. Abrahão and G. Bressan, “Video quality metric for streaming service using DASH standard,” IEEE Trans. Broadcasting, vol. vol. 62, no. 3, pp. 628-639, Sep. 2016.
• [13] T. Hoßfeld, M. Seufert, C. Sieber and T. Zinner, “Assessing effect sizes of influence factors towards a QoE model for HTTP adaptive streaming,” in 6th Int. Workshop Qual. Multimedia Exper. (QoMEX), Sep. 2014.
• [14] S. Afzal, R. Prodan, C. Timmerer, “Green Video Streaming: Challenges and Opportunities.” ACM SIGMultimedia Records, Jan. 2023.