The Greening of Streaming
Streaming video is wildly popular. But more than that, streaming is necessary, fundamental to the way international business is done and global entertainment is distributed, and—as we write this article in early April 2020, during the height of the COVID-19 crisis—the most efficient and effective way to foster connections across international borders. But does that efficiency extend to energy use? And, more broadly, what impact is streaming having on the environment?
These are questions that our editor, Eric Schumacher-Rasmussen, had asked us to contemplate months before the general public had heard terms like COVID-19 or social distancing or read hashtags like #saferathome in their Facebook, LinkedIn, or Twitter feeds.
And while it may seem foolhardy to discuss environmental impact during a time when there's been a massive spike in streaming usage—not just in the one-to-many way we typically think of it, but also in the many-to-many
use cases in which soft-client video chat and videoconferencing have kept small business engaged with its isolated clientele and grandparents in touch with their grandchildren, even just a few blocks away—this article is going to do just that. We'll delve into the complexities of streaming workflows to consider whether there's a mutual path forward for both streaming innovation and streaming efficiency.
Making the Case for Green Streaming
Co-author Dom Robinson, who created the GreeningofStreaming.org site, wrote a blog post laying out some of the environmental impacts of streaming, both current and near term. To date, there have been limited research efforts into understanding power consumption issues, and most focused solely on data center impact.
The impact is much larger than just what's stored at data centers. As Robinson points out in his blog post, co-author Tim Siglin did a study a few years ago on DVRs and observed that DVR hard drives were kept constantly spinning and recording—even in sleep mode—in anticipation of being ready in an instant with content the consumer would probably want to watch. It turned out, though, that almost 90% of the time, the consumer didn't want to catch up with all of the content that had been DVR'ed, meaning that all of that effort and energy were wasted.
The same could be said for streaming recommendation algorithms that pre-cache content based on some semblance of popularity: The effort and energy to create "instant on" functionality may be consuming significant amounts of power. We don't know for certain, however, since there have not been measurements, test beds, or concerted research attempts to delve into the holistic issues surrounding streaming from encoding to delivery.
There are enough discrepancies in the research that has been done so far to bring some of the results into question. When The Shift Project's "Climate Crisis: The Unsustainable Use of Online Video" report came out in 2019, a number of fingers pointed at the most obvious target: Netflix (this was before the launch of Disney+). Greenpeace even started the #ClickClean campaign to make Netflix and other platforms more transparent about their power consumption.
George Kamiya, digital/energy analyst at the International Energy Agency (IEA), wrote a compelling response to The Shift Project's report, noting that some of the figures in the report were questionable and giving solid grounding for counterarguments without denying the reality that streaming video causes significant demand for power. While both The Shift Project and IEA share an interest in shedding more light on the issue, a conversation with Kamiya made two points clear: There's 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 such as The Shift Project's.
In particular, there's been a tendency to scapegoat Netflix, which may explain why the company withdrew from invites to discuss the issue publicly. It is important to note that Netflix clearly and indirectly drives a vast demand for underlying energy. But as a significantly virtualized entity (operating numerous services within the physical infrastructure of ISPs that, for example, host their Open Connect caches, or where encoding and origination services are provided within Amazon's EC2 infrastructure and cloud), it likely has a fairly small electricity bill of its own, and so it's misguided to focus blame on the service.
Most of the research on streaming power consumption and environmental impact has focused on data centers, but their effect is only one piece of the puzzle.
The general media's lack of structural understanding of how streaming video workflows operate is fairly endemic, yet we have failed to drive the conversation ourselves in the trade press and at industry conferences. Indeed, in 10 years chairing Content Delivery World in London, and longer as a participant in (and recently chairing) the Content Delivery Summit in the U.S., Robinson notes that he has not once had a discussion about end-to-end power efficiency and sustainability of content delivery workflows. (You can expect that to change at this year's Content Delivery Summit at Streaming Media West.) There has been talk about the power efficiencies of GPUs, CPUs, and codecs, but these discussions have been myopic and not holistic. There is often an assumption that these challenges are someone else's responsibility, but as the architects of these very end-to-end systems, we need to place this responsibility on our own shoulders. We can make significant differences, and as technologists and architects, we don't even need to ask permission—just tell the corporate side that it is their only option. As the saying goes, "The geeks have inherited the earth." Now we must make it sustainable.
Even more than just inadequate reporting and research, though, the case against the greening of streaming seems to rely on the assumption that we're better off now than we were a decade ago and that streaming has created a net positive impact in both convenience and energy efficiency. At least we're not shipping DVDs around anymore, so we must be better off now than we were before.
While it's true that there are fewer shiny-but-scratched discs in landfills these days, the carbon offset of streaming may not be as significant as one might think. After all, it's not like the mail carriers that transported Netflix DVDs have stopped driving their routes. Nor have the number of living room television and DVD players decreased; rather, streaming set-top boxes have been added alongside the DVD and gaming console. Smartphone ownership has dramatically increased, and with it, both the need to power these media-capable phones as well as the "planned obsolescence" of older smartphones that consumers must now replace with newer smartphones that have chipsets to tackle the latest video compression.
Over the coming year, both Robinson (via his company id3as and GreeningofStreaming.org) and Siglin (via the Help Me! Stream research foundation) look to move the discussion forward and respond to the naysayers, with help from industry publications like Streaming Media and streaming media experts like you. To establish the parameters of what we hope will be an ongoing conversation, let's start by looking at steps throughout the unidirectional encoding and delivery process, considering a few areas where the industry might be able to tweak energy requirements.
Encoding (Ingest and the First Mile)
We've heard a lot about newer compression technologies, but a fundamental issue remains: Encoding content requires a significant amount of energy. Consider that it takes multiple years to move from software encoding to integrated hardware encoding—exemplified by H.264 and, more recently, by the fact that late 2019 was the first time that AV1 hardware-assisted encoding became available—and it's clear that the balance between innovation and power efficiency needs to be contemplated at the very beginning of the streaming media workflow.
Typically, "fixed encoders" are found in quantity in IP cameras and CCTV applications. These are sometimes ASIC-based, and at other times, they are system on chip-based.
The more the technologies become standardized, the more useful they are for law enforcement (submitting evidence from CCTV has a complex set of requirements), and, therefore, progression toward a globally fixed set of video compression strategies/codecs/protocols helps the production move to ASIC encoding. And in the scale of hundreds of millions of devices worldwide, reducing the power demand even 10% can have a significant impact on the global energy demand.
If field encoders' power usage can be significantly reduced, then renewable/solar charges can be put to use even more effectively. In our experience, a typical portable 8-hour solar charger today can power a small CCTV for 4–8 hours. We've also been experimenting with products like the PowerFilm flexible solar panel-battery bank combination and see promise for using its direct 12-volt barrel connector to power single-board computers (SBCs) that have the same connector type, without the need to convert to USB power connectors, which can be a bit unreliable for constant field use.
As the power demand for encoding is reduced, the ratio increases. If we ever reach the ideal situation in which the solar charger can charge and power the camera for 24 hours, then every outdoor CCTV camera could be removed from the grid. Given that at night, CCTV in schools, public buildings, and so on is one of the most common demands on "empty property" power supplies, this should be a key target for sustainability in the IP streaming sector—albeit largely forgotten compared to the higher profile but significantly lower volume of high-end broadcast streaming encoders used for OTT content.
Wireless or Wireline?
There is a complex debate about the relative power efficiency of Wi-Fi versus Ethernet, particularly in the less-than-1Gbps domestic and small and medium enterprise (SME) context. Most agree that Ethernet is likely more efficient (as witnessed by a debate on Reddit last year.
Wi-Fi networks are generally prone to packet loss and retransmission, and that inefficiency itself is reflected in more power demand at both ends of the link. In addition, if devices are on a wireless network but plugged in to power, there's the inefficiency of power consumption of both the Wi-Fi and the device itself, whereas Ethernet tends to pull lower overall power and guarantee a more consistent packet delivery approach.
Data Center (Storage)
The one area in the streaming and content delivery ecosystem that has a long-running focus on power efficiency is the data center. But while reducing the demand within the data center is crucial, it is never as easy as simply reducing the power consumption; the source of that power is also critical.
Many renewable energy sources do not provide a constant round-the-clock supply, and for the 24/7 global internet, this presents challenges. This is why Google, Apple, Facebook, and others locate their data centers in North Carolina. While all of these companies are commendable in their pursuit of sustainable/renewable power strategies, it is the easy access to Virginia's Appalachian mountaintop removal coal mining that may be one of the key reasons for their location choice.
While leading companies like Google, Apple, Facebook, and others are commendable in their pursuit of sustainable power strategies, they still place a high demand on coal power (and therefore coal mining).
The fact that these companies can place a continuous high demand on coal has shored up the coal-mining infrastructure and lowered costs to both the companies and the general public, which also benefits from the reduced cost of domestic energy, according to the documentary film Data Center: The True Cost of the Internet. There are more than 500 such mines in operation, and in addition to the complex issues of carbon release by burning the fuel, each of those mountains' ecologies have collapsed completely or are in the process of doing so.