Practical Approaches to Sustainable Streaming
Three years ago this month, the streaming industry woke up to the need for a greener approach to streaming. In the years that followed, rather than just pay lip service to environmentally friendly streaming, a number of companies and organisations are actively working to craft solutions that ensure sustainable streaming will lower overall energy and its corresponding environmental impact.
This article highlights where we’ve come from—looking back at live-linear delivery over the past decade, as well as more recent moves to lower power consumption—and where we’re heading.
We’ll do so by highlighting key practical areas that working groups of a relatively new industry organisation are using to shape the dialogue around various parts of the streaming workflow, from encoding to delivery to end-user device recommendations.
For those who might not know, the organisation Greening of Streaming (GoS) that Dom Robinson founded borrows its name from a watershed article that he and I wrote for Streaming Media magazine in May 2020. In that article, we advocated for both industry-wide discussions and actions to create streaming media delivery solutions that accommodate not just pricing and performance—two key factors in the rise of streaming as a delivery medium for both on-demand and live-linear content distribution—but also solutions that are purpose-built for a third p, power.
GoS and its Working Groups
The GoS organisation comprises many of the most significant operators, service providers, and technology vendors in the streaming industry. It acts to improve energy efficiency and sustainability efforts relating to streaming services architecture and design, promoting power as an equally important design consideration to price and performance in system development.
As a sustainable-media industry consortium founded in 2021, Greening of Streaming was recently awarded an inaugural 2023 NAB Show Excellence in Sustainability Award in the Leadership category.
With many active international working groups—drawing equally from academic, non-governmental research and policy organisations (NGOs) as well as representation from almost all aspects of the streaming industry’s workflow—GoS is neither an accreditation nor offsetting group, nor is it a standards development organisation (SDO), although many of our members also sit on SDO committees in various industry standards organisations.
Current Greening of Streaming membership as of June 2023
At its core, GoS is an industry-led user group (UG) that seeks to influence—via public perception and ongoing research into best practices for stream acquisition and delivery—the premise that SDOs should encourage consideration of energy efficiency as a “first-class” key performance indicator (KPI) in the development of streaming-related technical standards.
And finally, of key interest in the streaming industry, one of the GoS initiatives, the Low-Energy Sustainable Streaming (LESS) accord, is both an actionable premise and a call for wider participation across the industry.
GoS invites members to propose sustainability initiatives to the LESS working group via an online form.
In this article I’ll expand on prior sustainability approaches, detail reasons to consider additional areas of power-consumption optimisation, and invite diverse stakeholders across streaming workflows to engage with Streaming Media magazine, GoS, and like-minded stakeholders in a mid- to late-2023 initiative around the LESS Accord, with an overall goal of readers recommending measurable best practices for consideration by September 2023.
Prior Power Consumption Initiatives
The premise of sustainability in media delivery has scientific merit spanning back a dozen years, with founding GoS members having been involved in fundamental research on this topic for more than a decade. Yet for all the research done in those early day, progress is often slow.
Here’s an example of that slow progress: One of these research projects centered on legacy on-demand content distribution for MVPD, over-the-air (OTA), and direct-to-home (DTH) distribution, centering on the significant power consumption of in-home digital video recorders (DVRs).
That research formed the basis of the industry’s 2012 Base Case of estimated DVR consumption that is still in use today, showing that in-home DVRs consumed the estimated power equivalent of three nuclear power stations in the continental United States alone based on the fact that most in-home DVRs consumed 33 watts of energy per hour when in use (an average of 14 hours per day for 169 kwh per year) with an almost equal amount of power (31 watts) consumed during the 10-hour sleep cycle.
The primary difference between active and sleep modes was a dimming of the DVR-equipped device’s screen. The hard drives kept spinning—and recording mostly unwatched content, which research found was abandoned the majority of the time by consumers, who returned to their televisions to watch something other than the “catch up” content had been just recorded—even if the consumer was not present in the home.
The reason for this higher-power-consumption approach was a flawed use-case assumption. Overall industry feeling was that consumers would not wait the 3?4 seconds required for a device to “wake up” from an actual sleep mode and spin up the hard drives to access previously recorded content. So the hard drives that recorded DVR content never really stopped, and subsequently the power consumption level never dropped to anywhere near a true sleep cycle of 3?4 watts.
Even so, the findings were a significant wakeup call that led to a voluntary industry agreement intended to address power consumption that “covers the set-top boxes that consumers obtain from cable, satellite, or telephone companies that service more than 90% of the U.S. pay TV market.” First signed in 2013, the Set-Top Box Voluntary Agreement was iterated in 2018 and extended again last year to cover device power consumption until 2025.
Attempts to Self-Regulate During Growth
A number of companies signed the Set-Top Box Voluntary Agreement in 2013, stating the intent to lower overall power consumption below the 2012 baseline. But the approaches the industry took from 2013?2019 fundamentally missed a key research recommendation: move all media delivery—from on-demand to live-linear to live-event broadcasts—to IP-based streaming devices.
Set-top box voluntary agreement signatories
For instance, one approach the industry took was to accelerate the rollout of digital terminal alters (DTAs) that consume a lower overall wattage than the main DVR unit, delivering a television signal to a household TV using 5 watts in active mode and 4 watts in sleep mode. By contrast, the third-generation Apple TV was available at the time of initial research in Autumn 2012 and consumed less than 2 watts of power in active mode.
Worse, though, DTAs were reliant on a central DVR-equipped device to orchestrate delivery of content to each DTA. meaning the less power-consumptive units were only rolled out in secondary rooms (the “spare room TV” in consumer lingo). So while the averaging of power consumption across all cable- or satellite-delivered set-top box devices in a home declined, the actual power consumption in that same home did not decline.
This goes to a fundamental stance that Robinson espoused as he launched the GoS organisation: no greenwashing.
“The first and most important rule for Greening of Streaming members and partners is no greenwashing,” said Robinson in a recent interview. “This has been a powerful impetus to truly solve power consumption problems across the media workflow. We’re also peer reviewing each other to make sure that our marketing doesn’t make claims that are unsubstantiated just to try to win competitive sustainability points, which aren’t really—when you look at the big picture—sustainable.”
Fast-forward to 2021 and details that emerged in the ninth report on power consumption levels from the initial Set-Top Box Voluntary Agreement. Using self-reported numbers from the Voluntary Agreement signatory companies, NCTA reports that power consumption rates for set-top boxes have been cut by 57% from the 2012 base case of 33 watts. NCTA goes on to report that the power savings equate to an estimated $2.5 billion in consumer energy cost savings in 2021 and approximately $12 billion overall since the Voluntary Agreement’s inception.
How did they get to these 57% overall power reduction levels? By finally moving to IP-based media delivery devices.
“A major driver of savings under the program is a transition from traditional digital video recorders (DVRs) that include hard drives to non-DVR Internet Protocol devices,” NCTA reports, adding that “these IP devices represented more than 70% of all purchases in 2021 and on average use only 36 kW hours per year, compared to an annual average of 267 kWh for DVRs in 2012 when the agreement began.”
Can We Go Further?
The voluntary industry actions taken a decade ago to reduce power consumption for media delivery—from the initial set-top box Voluntary Agreement to the more recent Small Network Equipment (SNE) that also covers “residential broadband Internet access services” and devices that, in turn, facilitate streaming media delivery—are quite commendable, but I contend there’s much more to be done, both at the consumer and infrastructure level.
Readers might ask about the benefits of incremental gains at a consumer level, especially with flash-memory-based IP-delivery devices bringing active use wattages down to a much more reasonable level and offering a true sleep cycle that often drops power consumption below one watt. We’ll address those in a later section of this article.
At the IP-based delivery infrastructure level, however, our premise is that an underlying flaw in the design of delivery systems has the potential to dwarf consumer power consumption savings, in no small part due to the streaming industry’s initial focus on price and performance.
“One of the premises we put forward was this idea that certain parts of the industry, such as the content delivery networks (CDNs), spent the first ten to fifteen years of its life focused on price,” said Robinson in another recent interview. “We think it’s all about performance, because in the more recent decade it’s all been about performance, but the first decade was all about making it affordable to you to get high-capacity content or high-data rate content."
The price focus was necessary in the first decade of streaming (1998?2007), so that performance increases in the second decade (2008?2017) could be afforded as streaming delivery moved to scale.
On the performance front, however, there appears to be a fundamental disconnect between power consumption during actual broadcasts compared to two other stages, which I’ll call the prepping-for-broadcast and idle-time stages.
Robinson also started GoS with a bit of what he calls a “null hypothesis” where power consumption in the delivery infrastructure does not significantly vary between these stages, because delivery networks are maintained at close to peak capacity and at potentially overly redundant high-availability levels that are unnecessarily power consumptive.
Testing the Premise
To test that hypothesis, several GoS working groups have been formed to measure power consumption across the entire acquisition, delivery and consumption ecosystem. For instance, one working group is measuring power consumption across a number of member CDNs, both before and during peak global live-streaming events. Another working group is recommending an initial key performance indicator (KPI) to be used during the creation of both end-user equipment—we’ll reference it for the remainder of this article as customer premise equipment (CPE)—as well as media delivery infrastructure gear.
GoS also engaged with various organisations, from EBU, IBC, and NAB on the broadcast front to streaming technical alliances, such as IETF, SVTA, and the newly formed CDN Alliance. The intent is to both better understand the role of power in peak-versus-non-peak media delivery periods as well as propose standards frameworks around the use of streaming codecs, CPE and infrastructure solutions.
CDN representatives note, rightly, that they’re acutely aware of power consumption each month as they pay power bills, so it is in their best interest to balance customer-consumer demand while also maintaining a reasonable rate of power consumption. To be clear, we’re not advocating for an “always off” mode, but it’s a bit ironic that today’s IP-based media infrastructure is somewhat akin to the “almost-always-on” approach the industry took to “sleep mode” for the DVR-equipped devices mentioned in the prior section.
To illustrate the vastness of the streaming infrastructure—as well as the relative size of the power consumption portion affected by the consumer portion of delivery workflow—this figure illustrates the end-to-end workflow in four sections. Click the image to see it at full size.
When is LESS More?
Earlier I mentioned the fact that CPE equipment has moved towards lower overall power consumption, primarily as part of a move away from hard-disk-equipped DVRs and into a broader uptake in IP-addressable CPE equipment. Here’s where that plays into current efforts.
The premise from GoS on this front is that the consumer devices have a power range that’s directly affected by the content delivered to them.
Let’s use the example of a flat-panel LED television to illustrate this point, based on data that Robinson measured with a watt-meter mains-power device against his TCL 65RP620K Roku 65" Smart 4K Ultra HD HDR LED TV.
While an electronic programming guide (EPG) consumes approximately 75w, setting the EPG as a baseline for overall consumption, the difference between standard-definition (SD) and high-definition (HD) viewing at standard frame rates only rises slightly to 85w. The use of high-dynamic range jumps quite significantly to 135w.
In the same way, the various permutations of codec, frame rate, resolution, and colour bit depth have a significant impact on not just the power consumed by CPE equipment but also on the broader compute, transit and delivery infrastructures themselves.
That’s where the Greening of Streaming new LESS Accord (Low-Energy Sustainable Streaming) comes into play.
I posit that, in many cases, consumers cannot differentiate between various streaming and broadcast service qualities. Indeed, the industry itself increasingly relies on computer-aided quality techniques to differentiate “acceptable” and “unacceptable” levels that even the humans programming these algorithms and monitoring the output cannot perceive.
The LESS Accord aims to dig deep into the heart of the broadcast and streaming industry and ask a taboo question of a quality-obsessed industry: Can our industry collectively deliver a lower-energy equivalent of current high-quality content, thereby lowering overall power consumption across CPE and infrastructure equipment?
To do so, broad industry participation in both concepts and testing stages is needed, so a key motivator behind the LESS Accord is to “give permission” to stakeholders to ask out loud what many engineers in the industry already instinctively, privately think. By exploring ways in which the industry might be able to deliver services that fulfill consumer’s visual and auditory quality expectations, it’s rational to conclude that the resultant reduction in energy demands at various parts of the streaming workflow may have net positive environmental and economic impacts.
As the LESS Accord seeks to reduce power consumption, I implore the industry to avoid an easy trap: simply overselling imperceptible quality/value propositions which yield inappropriate, expensive, unsustainable, and unnecessary energy demands with no perceptible benefit to the viewer.
The timeline for initial LESS Accord engagement is meant to drive rapid innovation through the following steps undertaken by GoS members and partner organisations:
- Work to identify companies or individuals that have viable and innovative LESS ideas, completing this task by early summer 2023.
- Define real-world production testing models, based on these best-practice concepts by autumn 2023. GoS intends to present these testing models at IBC in September 2023
- Run tests against these models throughout Q4 2023 and Q1 2024.
- After initial testing of each best-practice premise is complete, work through Q2 2024 to produce outcomes and a proposed accord at the end of that cycle.
As this article wraps up, let’s also quickly recap what working groups are handling the various tasks that led to the LESS accord concept. To build on the Working Group 6 efforts, which Sam Orton-Jay has led through the initial LESS Accord definition phase, GoS is in the process of forming Working Group 8 to focus on approaches to CPE measurement.
This new effort will also intersect with Working Group 4’s efforts to measure power consumption on the delivery infrastructure mentioned earlier in the article, as well as Working Group 7’s efforts around KPI initiatives for building more sustainable delivery and CPE solutions.
Scaling appropriately dovetails into power considerations for both CPE efforts such as the LESS Accord and the two Voluntary Agreements mentioned at the outset, as well as into the cloud-compute infrastructures which are often required to be ready at a moment’s notice to scale up to meet demand.
The end result of that always-ready state means the majority of the equipment used in cloud-compute instances must continue to run near peak capacity for extended periods during which there is no corresponding scaled delivery requirement.
To more effectively balance power against performance, this must change. Doing so starts with a mindset around everything from major sporting events to free ad-supported (FAST) television as illustrated in the following quote:
“We’re so often focused on the Super Bowl or the Olympics, or we’re focused on Netflix and Disney,” said Eric Schumacher-Rasmussen, Streaming Media conference chair, at a recent industry event. “The vast majority of companies that are streaming are not reaching anywhere near those numbers, nor will they ever or have any need to. This [scaling expectation] then creates an imbalance that if it’s really costly and really hard to create a FAST channel, you won’t be satisfied with the revenues that come in since the viewership may not be there.”
Setting these expectations appropriately on the three p’s—price, performance, power consumption—may also have a positive net price effect, as the cloud-compute and CDN companies will have flexibility in their power consumption schemes, a portion of which may be passed along to their customers. In other words, it’s equally about making the industry profitable and saving the planet. With efforts such as the LESS accord, I believe it’s possible to do both for an entertaining-yet-sustainable future.