By Sam Orton-Jay, VP Product Marketing, V-Nova
Whilst short of the 7.6% estimated reduction needed to keep global temperatures within 1.5ºC of pre-industrial levels, if these reductions could at least in part be maintained through a continued change in attitudes towards working from home and remote collaboration generally, it could help to have a meaningful impact in the fight to control climate change over the longer term.
Many have celebrated the lower ecological impact of less commuting and business travel but few have considered the increasing impact of the exponential growth in internet streaming. These changes have put tremendous new pressure on our video delivery systems whether it be video-conferencing or entertainment services of many kinds. One clear example of that was how Zoom went from 10 million daily meeting participants early in 2020 to 300 million by October. It has also shone a spotlight on the environmental impact of our video services and the many data centers which ultimately power them.
So, what are the facts of the situation? Is video really just a tiny side issue in the world of carbon emissions or something we really need to look at carefully amid all this growth in consumption? The International Energy Agency (IEA) says streaming services from videos and gaming are driving demand for data center services, and that they will account for 87% of internet traffic by 2022. Those data centers that process and store data from online activities, such as emailing and video streaming, accounted for about 1% of global electricity use in 2019. Encouragingly, despite a 60% surge in demand for data center services, the IEA believe the energy required to power this will remain flat until 2022, thanks to strong investment in energy efficiency in the space. However, as those quick wins in efficiency are exhausted and the growth in video consumption continues apace the energy efficiency of our video services is set to only become more important to ensure that the ecological gains we make from remote working are maximized.
This brings us to the core question of what can our video services do to reduce energy consumption?
Of course, there are other factors driving the priorities of these services. Entertainment viewers are demanding ever higher resolution and quality streams that take advantage of the steady increase in pixels our TVs and other devices have today. That is pushing services towards using more recent and advanced compression technologies like HEVC and soon emerging codecs like AV1 and VVC. These can all help to deliver bigger and better pictures within realistic bandwidths to our homes whilst also helping to mitigate against the other impact of the pandemic, maintaining reliable services despite heavily congested networks as a result of the increased traffic (it rose by 40% in just the first 3 months from February 2020). However, these technologies come at great cost to the ecological goals we’ve discussed here. As a rough guide each generation of compression technology adds a 10-fold increase in the processing complexity of encoding video. That directly correlates to power consumption in the data centers doing this work. Encoding is only one step in a typical delivery workflow, but it is by far and away the most compute intensive and therefore accounts for a large proportion of the energy requirements.
Thankfully, another emerging industry standard is set to contribute to mitigating against these increasing compute costs: MPEG-5 LCEVC “Low complexity enhancement video coding” is unique in the world of compression standards being an enhancement to all other codecs. In addition to improving their compression efficiency (quality at any given bitrate) possibly more importantly it can reduce the processing cost of these newer compute-intensive codecs by up to 75%. If used in conjunction with both the ubiquitous AVC/H.264 and codecs like HEVC and AV1 it could have a profound impact on the overall emissions of video everywhere.
Whilst video encoding is undoubtedly a critical component, further research is really needed to fully understand the relative energy consumption of different parts of the workflow and for different types of service. For example, Netflix is an often-cited example but does relatively little encoding since it has quite a small catalog of content when compared to something like YouTube which is encoding thousands of videos each minute. In the case of Netflix the bulk of its energy impact is quite probably in the internet transmission of millions of subscribers playing the same streams. Those costs are quite well understood with the electricity intensity of fixed-line internet data transmission (recently calculated at 0.015kWh/GB) and can be used to evaluate the impact. Indeed, those values were recently used in an analysis of video-conferencing provider Zoom’s energy impact but that notably omitted any consideration of the additional power consumption on our devices of the video encoding and decoding that the Zoom software employs simply because it’s difficult to calculate.
In summary, our digital video world represents at least 1% of global emissions even by the most conservative of estimates and for that reason is undoubtedly an important frontier in our collective push for sustainability. As consumption continues to grow and more energy intensive technologies like new codecs are employed to achieve bandwidth efficiencies, services must prioritize solutions to mitigate against these power increases like employing MPEG-5 LCEVC. Moreover, video services of all kinds need to work with technology vendors and even the academic community to significantly improve our understanding of the energy costs of every component in their delivery workflows from encoding and internet transmission all the way through to the rendering on our screens. Better profiling will help us to identify where the key problems lie and drive the creation of effective solutions to the problem.