The State of Streaming Sustainability 2024
Every year, there seems to be a new power-hungry use case for computing applications. A few years ago, it was cryptocurrency. This year, it’s machine learning (being sold as AI, which doesn’t actually exist, as simple gaming of machine learning algorithms veering easily toward screed and hate language has shown).
In all of the hype around crypto and AI, one might assume that streaming doesn’t really consume enough energy to significantly move the needle on power generation and consumption. Yet, while there are numerous narratives floating around on the topic of streaming’s power problem—a few of which I’ll cover in this article—there’s a bigger issue at play with streaming: How do we make it sustainable, even as it grows to be the dominant form of audio and video delivery globally?
Cutting Through the Hype
Several simplified metrics have been applied to the power consumption discussion around streaming, many of which are phrased in carbon emissions language. The most persistent of these metrics was a byproduct from a 2019 report on the impact of streaming from The Shift Project in France. After the release of the report, news organizations noted that the carbon emissions (CO2) from watching 30 minutes of Netflix was equivalent to driving a car approximately 6 kilometers (or roughly 3.9 miles).
In a 2020 follow-up, The Shift Project said that a third-party analysis of its report by George Kamiya was “the most comprehensive and transparent review of our 2019 results” and agreed that his commentary, later posted on the IEA’s website, “does point out … a figure that is undeniably wrong.” The organization tried to justify its models by stating that data centers consumed more power than had been initially projected.
Despite the fact that The Shift Project revised its numbers by a factor of 8 (having miscalculated bits and bytes in the initial report), noting that “there was indeed a major error on the bitrate”—3 megabytes per second instead of 3 megabits per second—“which accounts for 90% of the discrepancies exposed,” and blaming the analogy on a collaborator, the 30 minute/6 kilometer metric stuck along with a figure of 1% of global greenhouse gases being produced by streaming.
So with the 2020 correction to the 2019 report by The Shift Project—which has the organization stating that “after correction of the bitrate (bit-to-byte conversion error, i.e., a division by 8 of the bitrate), our model indicates 0.77 kWh per hour of video”— it would be safe to assume that the baseline metrics would also be corrected. Unfortunately, even in 2023, the 30 minute/6 kilometer metric was still being reported as accurate by academics, news organizations, and even the general public.
“Streaming media is calculated to contribute a surprising 1% of global greenhouse gases,” according to a team from Simon Fraser University, adding that the information and communication technologies (ICT) market segment—of which the team and others consider streaming media a subset—“is responsible for 3.3% to 3.8% of global greenhouse gases.”
Speaking of ICT, even more recent 2023 writings around environmental impacts of streaming still try to use the flawed Shift Project data as justification for claims that “streaming media is killing the planet.” “ICT engineers have strong agreement about the electricity consumption of streaming video,” according to the Simon Fraser University team. “Their findings confirm The Shift Project’s estimate that streaming video is responsible for over 1% of global greenhouse gas emissions and growing fast.” Except there’s no consensus among ICT engineers, even those in the streaming industry, perhaps not only because of The Shift Project’s errant data, but also because of the ongoing efficiency efforts led by the industry itself.
Defining the Problem
Even though there’s no consensus, that doesn’t mean there’s no problem. Streaming viewership continues to climb, even as cryptocurrency mines are decreasing. Live-event streaming viewership is climbing even faster, blunting some of the gains made by caching on-demand content during off-peak hours as a means of increasing energy efficiency.
Back in 2020, before The Shift Project corrected its data, even those of us at Streaming Media fell for the easily reported 30 minute/6 kilometer metric in our first The Greening of Streaming article. But we also noted the dissent in the ranks around the data (George Kamiya) and that the broader implications were that there was “disagreement on how to measure and interpret the data around sustainability and energy efficiency, and the popular media outlets and politically minded organizations are ill-prepared to deal with the nuances and are likely to sensationalize” reports like the one from The Shift Project.
In subsequent years, champions of streaming sustainability have not been limited to the Greening of Streaming organization, as the Streaming Video Technology Alliance and the newer CDN Alliance have entered the fray as well.
One of the major issues in all of the discussions around sustainability is a lack of common lexicography and, perhaps to a greater extent, a common measurement methodology. Greening of Streaming has a working group (WG1) dedicated to defining the lexicon, and the goal is to produce accurate measurements and methodologies. Building a bit on disparate definitions across a number of academic institutions —such as the University of Bristol, Simon Fraser University, etc.—as well as those from NGOs and research organizations, these definitions are critical to characterizing the problem.
Once the definitions are in place, methodologies that build on prior research from small businesses (Radiant Media, Touchstream), research organizations (DTG, Help Me Stream Research Foundation), and large companies (Akamai, including the former Lumen team) can be further honed to generate aggregated efficiencies to address streaming’s unique challenges for scaling live-event delivery.
In the meantime, the industry is actively working to define issues that can be addressed in isolation. One example of this comes from a 2023 blog post from Brenton Ough, CEO and co-founder of Touchstream, that’s centered on analyzing and monitoring streaming media delivery. “The demand for streaming had a ripple effect: streaming operators needed more infrastructure to run longer and, as they pushed their workflows into the cloud, cloud operators needed more infrastructure to support the demand,” he writes. “All of this has to be monitored,” he continues, noting, “Probes that are continually monitoring streaming components, even when there isn’t sustained demand, are wasting energy.”
Worse, the data that’s collected might not actually impact the QoE that monitoring is geared toward delivering. So, if measurement solutions run unabated and aren’t positively impacting QoE for the end user, how do we make sure the monitoring itself isn’t adding to the energy consumption that we’re trying to reduce?
Calls to Action
One of the calls to action that Ough points out is a practical and beneficial implementation of machine learning. He notes that “a smarter monitoring system can reduce unneeded cycles and help the monitoring activities contribute to improved sustainability across the workflow. The first step is to programmatically align probe activity with audience scale. As scale decreases, the probes deliver data less frequently, requiring fewer cycles and reducing the demand for CPU or GPU usage.”
Another step in this call to action around monitoring is to use machine learning to predict potential spikes in delivery—and the need for additional monitoring—based on past trends. Note, though, that past results don’t guarantee future results.
A second call to action runs along the lines of limiting stream lengths or file sizes. An example of this is the Small File Media Festival. The fourth iteration was held in October 2023 in Vancouver, Canada, with more than 60 short films displayed at a local arthouse theater, The Cinematheque. The festival’s webpage notes that the films “are small in file size, but huge in impact: by embracing the aesthetics of compression and low resolution (glitchiness, noise, pixelation), they lay the groundwork for a new experimental film movement in the digital age.”
So how small is small? The criterion for the festival is that content should be “no more than 1.44 megabytes per minute, the storage size of a floppy disk.” Yes, you read that right: a floppy disk. The festival’s webpage notes that “TikTok is positively bloated in comparison, at 70 megabytes per minute.” Examples of file sizes for the 2023 festival can be seen by scrolling down here.
“Watching small-file media together on a big screen brings the democratic potential of cinema into the digital age by showcasing artworks made with eco-friendly practices, affordable equipment, and minimal processing time (all without sacrificing artistry or immersion). Small-file creators use ingenious techniques to make these tiny movies beautiful and effective,” according to the festival’s webpage.
The concept behind smaller files goes not only to the technical sustainability of small-file delivery but also to the overarching impact that intermittency has on media delivery. The festival’s webpage couches this in terms of the workers’ struggle (“bandwidth imperialism,” “political sustainability,” and “state suppression that takes the form of internet blockade and restriction”), but the festival organizers are also not opposed to streaming the content itself to reach a wider audience (via Vimeo’s showcase platform).
A third call to action centers on the actual compression used for newer devices. “Two things can help in the long-term to get the highest possible efficiency,” writes Remi Beaudouin, chief strategy officer at ATEME, a founding member of Greening of Streaming, in a 2022 Streaming Media post. “One is the introduction of new codecs; the second is encoders using standards-compliant algorithms. Both can work together to reduce bitrate and bandwidth consumption. This will lead to a lower carbon footprint as less storage, streaming capability and caching is required.”
Noting that deploying new codecs requires new devices, Beaudouin concedes that this approach will potentially raise carbon emissions in terms of “carbon embodiment” that results from manufacturing these new devices. Yet, in the 2 years since Beaudouin wrote about newer codecs, the industry has begun to shift in that direction. More on this topic can be found in panels from Streaming Media Connect 2024.
A fourth call to action comes from the programming side of ICT. In an August 2023 posting titled Green Streaming: A Guide to Sustainable Development, the team at Better Software Group (BSG) points out several ways to approach sustainable app development. One area where they argue that power consumption can be lessened has to do with the design of user interfaces in streaming applications.
"Crafting simple user pathways and prioritizing functionality with digital solutions play a pivotal role in sustainable and conscious development,” the BSG team writes. “By cutting back on excess resource-intensive elements like graphics or animations we can create not only easier-to-navigate apps but reduce device power consumption.”
BSG also notes, however, that there are trade-offs to be made: “Obviously, sustainability is not the only achievement you expect from your streaming solution.” But the trade-offs are worth the effort.
A final call to action centers on lean-forward experiences that bring viewers into an interactive stage with content. One example of this is an OTT service called WaterBear. While it’s a strange name, it’s perhaps not that much different from the names of early entries into the streaming space. The impact that WaterBear is attempting to make is certainly admirable.
Launched in an initial eight countries—Australia, Canada, Ireland, the Netherlands, New Zealand, South Africa, Australia, the U.K., and the U.S.— WaterBear’s goal is to engage viewers not only with compelling sustainability content but also with the NGOs working in those particular fields of expertise. “With original content and curated documentaries inspired by the United Nations Sustainable Development Goals (SDGs), WaterBear members will now be able to stream hours of video at any time, and on any device—all for free,” the company said in its launch press release.
Combining a group of celebrity ambassadors and key NGO partners (more than 100 at this writing) as a way to generate interest in its content, WaterBear also allows viewers to spread the word about projects they care most about. “WaterBear is not an entertainment platform as much as it’s a mix between entertainment, education, and action,” says Ellen Windemuth, the company’s CEO. “One of the perhaps most important things you can do while you’re watching [is that] you can engage with the NGO behind the project; you can donate to them, you can become a member, you can post things on social media help to amplify their message.”
In addition to offering ways to engage with the NGOs through donations and memberships, the WaterBear team makes decisions on how to best deliver content based on sustainability principles. According to the company’s website, “Streaming video content online, at scale, can be incredibly energy intensive. However, we are proud to have our videos streamed through Vimeo, which uses Akamai for their content delivery. Akamai has an impressive sustainability policy and is committed to reduce emissions and use renewable energy wherever possible. We consistently strive to make technological decisions that better the planet while maintaining the ultimate user experience.”
Conclusion
This brief overview of the state of sustainability doesn’t do justice to the continued work that’s being done by numerous engineers, researchers, and data center/ICT professionals working to drive down the overall power consumption of streaming. Yet with those errant 2019 datapoints still dominating the general streaming energy conversation in academia, it’s no wonder that general articles continue to espouse an apocalyptic vision of streaming’s growth. As recently as October 2023, a Utopia article cited streaming’s responsibility for “one percent of global CO2 emissions” as an attention grabber.
One can only hope that a newer set of data will emerge in 2024 to more accurately gauge true power consumption in the streaming industry. This will allow the industry to target further reductions in power consumption while working toward longer-term solutions that reuse old technologies alongside current best practices as a way to extend the life of streaming tech for years to come.
Related Articles
Streaming Media's Tim Siglin sits down with Greening of Streaming Founder Dom Robinson to discuss streaming sustainability and the LESS Accord in this exclusive interview from Streaming Media East 2023.
28 Jun 2023
So considering streaming, file transfer, or even leaving the long tail simply unused and archived as the storage media changes around it over years, there's much to think about for anyone wanting to leave a video as an intergenerational legacy.
21 Nov 2022
The time has long since passed to turn our attention to making the streaming delivery ecosystem more environmentally friendly. Here are some of the key challenges, as well as some suggested first steps towards solutions.
26 May 2020
Companies and Suppliers Mentioned