It's Not Easy Streaming Green
At Streaming Media East 2022 in Boston—our industry’s first in-person event in more than 2 years—I presented a session called Sustainable Streaming Lifecycles as a way to engage the industry in thinking beyond the present. Attendance was surprisingly good, especially since it was the last session of the last day, and the audience was engaged in the topic. This article both synopsizes that presentation and delves into other conversations that have happened since my first article focused on sustainability (co-authored by Dom Robinson), all in the context of keeping streaming’s global-scale ambitions in line with potential power-consumption challenges.
In my presentation, I started out by reminding the audience that, as an industry, streaming has been around for fewer than 25 years but has fundamentally altered the media-consumption landscape in a way that took traditional over-the-air (OTA) broadcasts almost 50 years to achieve.
Our first focus as an industry, dating back to the late 1990s, had a real technical bent. In those days, it wasn’t uncommon to hear questions such as “Can we do this?” or “Is it possible to even stream at scale?” After we passed through that first wave of self-doubt and proved that streaming was a viable way to deliver content—what we all called performance, powered by CDNs—it seemed the sky was the limit.
And then the bills started piling up. If you’ve been in the industry for some time, you’ll remember that we had everything from the 1999 Victoria’s Secret webcast to later live events that we were afraid would “break the internet.” It turned out that they didn’t break the internet, but they certainly broke the bank. Content producers for these live events got their CDN bill and were schooled in what’s called 95/5 or 95th percentile. Producers began to realize they couldn’t afford to deliver at scale, so we then had a number of years when we said, “OK, we can technically do this from a content delivery standpoint, but can we afford it?”
That led to another round of self-reflection, which happened to occur around the 2007–2009 recession, followed by several technical innovations (think HTTP-based delivery at scale) that led to another period of performance growth. What was absent from those conversations, though, was power. The consideration of power consumption—and long-term sustainability—never really measured up against the two juggernauts of price and performance, not at least until the outset of the pandemic. During that time, with lockdowns and the dual necessities of stay-at-home work and schooling, both price and performance took center stage, as scaling up of on-demand live events and live conferencing began to rapidly expand across the globe.
During the initial period of lockdowns, we sort of sacrificed power considerations in an “anything goes” effort to accommodate global scale, which can basically be summed up as, “Pay no attention to the number of power plants behind the screen.” Without conversations on this topic, though, there’s not a way to reach a sustainable model that balances global delivery (performance) at an adequate cost model (pricing) without sacrificing a significantly high percentage of natural resources (power).
Do We Need to Be Sustainable?
Some might wonder what the big deal is around sustainability and might even argue that power considerations will hold us back from achieving global television-sized audiences. The short response to this is that perhaps it will slow the growth curve, but it will also force the industry to innovate. The corollary, though, is what happens if we maintain the status quo for an industry that, by some estimates, uses almost 10 times the power of the overall cryptocurrency industry.
For the most part, however, companies in the industry are beginning to internally measure themselves against power consumption. This is, in part, due to the rising cost of energy, which has to be passed on to the end customer if profit margins are to be maintained. But it’s also become part of the conversation around how to serve customers who are concerned about overall environmental impact.
In the corporate sector, at least in the U.S., there’s also a move to have public companies measure their environmental, sustainability, and governance (ESG) goals. For companies that touch the streaming industry—including content owners, OVPs, CDNs, and cloud-based computing providers—the power-consumption question becomes much more pertinent if it means the potential of lowering an overall ESG score on an institutional investor’s investment scorecard.
One good example of the conversation around sustainability is taking place at the Streaming Video Alliance (SVA). I interviewed executive director Jason Thibeault at Streaming Media East 2022. “After you and Dom [Robinson] wrote your article for Streaming Media magazine, I thought, ‘We need to address this [at SVA],’ ” said Thibeault. “Then along comes [the organization] Greening of Streaming, and I kind of went, ‘Oh, somebody’s doing it, we don’t have to do that now.’ ”
When we wrote the article, nobody was really talking about sustainability, and we weren’t sure whether that would actually be a topic of interest to anybody. But it turns out it was a massively interesting topic to many in the industry, and subsequent conversations that we had with industry leaders led Robinson to found the U.K.-based organization Greening of Streaming, which recently brought on its first executive director and held its first public event at the U.K. Parliament.
Thibeault noted that SVA isn’t just relying on the Greening of Streaming organization to drive the conversation forward. “We’re getting ready to launch a sustainability microsite just for our members,” he said, adding that SVA companies will continue to talk about what needs to be done. “Having some Greening of Streaming content produced for us on a monthly basis [is key], as it is important to us to be involved at every point in the decision-making process when it comes to technology development.”
In other words, yes, we need to be sustainable, and that sustainability comes through collaborative industry innovation. More on that later.
How Do We Measure Sustainability?
During the 2021 Content Delivery Summit, in the run-up to the launch of Greening of Streaming’s membership drive, Robinson and I held a fireside chat with members of Akamai’s sustainability initiative. Robinson pointed out that the language itself is not yet set, which could impede accurate cross-industry measurements: “One of the things that’s complex as we start to see big public companies reporting on sustainability is the language isn’t yet settled, and we have different ways of measuring internally to how we might talk about that externally.”
Akamai’s director of sustainability, Mike Mattera, noted that many companies in our industry have to take into account measurements across vast global regions. “We’re very, very focused on the greenhouse gas emissions in the different areas,” he said. “We’re looking at things specifically focused on some more granular market-based emissions factors to be able to really kind of pull apart our overall emissions impact in these individual areas.”
Mattera went on to note that Akamai faces the added challenge of sheer scale. “Being in 4,200 locations and being across this kind of global spectrum, understanding that overall impact of the power that we’re using,” he said, “and the ancillary services like cooling and [uninterruptible power supplies] as an example, being able to factor those in and understanding the kind of the total impact really, really helps us measure, understand, and also factor into these overall emissions factors.”
Vinay Kanitkar, an Akamai Fellow who is also part of the company’s product and platform architecture team, talked about the challenges in designing and measuring sustainability. “At a broad level, the two important things for me are to use less power as we continue to scale,” he said, “and less power in terms of units of traffic/units of compute that our customers care about. The other side of it is to make sure that more and more of that power is coming from renewable sources.”
Robinson has noted in subsequent conversations that this two-pronged goal—using renewable energy and consuming less power—needs to emphasize the lower power consumption as the overarching goal. Otherwise, he argues, we could end up using renewable energy at much higher rates than necessary, rather than focusing on driving down overall power consumption as we scale streaming out to a global audience.
In addition, when it comes to the measurement of power consumption and environmental impact, a fundamental exception that Robinson and I took with the state of environmental discussions when we penned the “Greening of Streaming” article in early 2020 still needs to be addressed by the broader industry: Greenwashing through data attribution (essentially shifting the problem to some other portion of the streaming workflow) is unacceptable. As such, from the outset, the Greening of Streaming organization has taken a very strong stance: no greenwashing among member organizations.
Adam Curwin, who is Greening of Streaming’s first executive director, made this key point at the organization’s U.K. Parliament event in June 2022: “We have to address this challenge as an industry, not as individual companies.” He continued, “Say one of our members makes a local optimization in its video delivery. This might make their equipment more efficient and save 200 watts on one of their servers. Well done, very
green. However, if the impact of that change is that every device consuming the stream requires slightly more power to display the pictures, overall, you have achieved the exact opposite of the local intention. If each of a million TVs (each of which probably consumes around 200 watts individually) uses a few watts more, suddenly, your ‘green’ initiative is responsible for upping overall power consumption by several megawatts. This is hugely important and precisely why a cross-industry body is required to ensure joined-up thinking and spread best practices.”
At the Content Delivery Summit session, Kanitkar put an equally important weight on the overall picture of streaming at scale: innovation. “In fact, now I’ve come to the view that making things more efficient is harder than building new features and new capabilities. It’s innovation in a very different way; it’s innovation at a different level. And so we are investing a lot in that now.
“We used to do these big keynotes for Apple when Steve Jobs used to [introduce] the iPod and the iPhone and so on,” said Kanitkar. “And those used to be about 20GB of traffic … and we used to use hundreds of machines to serve 20GB of traffic. Right now, you could do that from one machine.”
Which isn’t to say the measurement of power is strictly a matter of measuring one server’s power consumption versus 20 servers. In fact, Kanitkar noted that Akamai is at “the early stages of trying things out in terms of FPGAs or ARM processors or alternate processing infrastructures.”
“We’ve used FPGAs and GPUs for some very specific applications,” said Kanitkar, “but across the broad infrastructure, even the 4,000 or so locations that [Mattera] mentioned, we’ve stuck to general purpose CPUs for the most part. … So we have come a long way in terms of general purpose compute … but [to innovate], we are looking at even more specialized hardware right now.”
The End of the Road? Or ‘What Do We Do With All of This Old Stuff?’
I covered the litany of special-purpose chipsets, such as ASICs, DSPs, and FPGAs, in a 2021 article and have to say that a properly used alternative processing infrastructure has the very real potential to lower overall power consumption while upending the price performance status quo, especially for repetitive tasks that need to be duplicated across thousands of global locations. But as the industry begins the shift toward balancing power consumption with performance and price, there will be an inevitable increase in the pace of replacing older gear with newer gear.
However, the advent of innovation and alternative processing begs a simple question: “In our pursuit of better efficiencies of scale, if we replace all of the old stuff with new stuff, what do we do with older technology when we’re done with it?” That’s a question at the heart of the Help Me Stream Research Foundation, the 501(c)(3) scientific charity a group of us started just prior to the pandemic.
The typical response to old gear is to recycle it. That’s not necessarily a bad approach, and, yet, there are ways to lengthen streaming equipment lifecycles, one piece of old gear at a time.
A major shift in thinking is needed, though, away from planned obsolescence and recycling as a competitive mindset. One hurdle to get over is to objectively acknowledge that, while recycling is a lucrative business, there are several key shortcomings to the use-and-recycle approach for computing gear. Recycling a server is a much more complex process than just melting down a plastic bottle into reusable pellets for the next soft drink bottle.
First, the recycling process for computing gear is primarily one of extracting precious metals from mainboards, power supplies, and storage units. And to extract these, oftentimes a number of rather harsh chemicals are used. What’s left over still has to be disposed of, and that byproduct can—by some estimates—equate to more than 90% of the actual gear’s volume that was made up of various types of metals and plastics.
Second, recycling means that usable gear is rendered unusable. If there’s anything the pandemic has taught us about the scaling of streaming offerings, it’s that consumer appetite for streaming continues to grow. In off-the-record conversations with representatives from several companies in the space, I learned that the use of older servers—which were set to be superseded by newer, more energy-efficient servers—had to be prolonged due to a combination of chip shortages, shipping delays, and the sheer scale of consumer demand.
Third, older technology still has significant life left in it. This goes to the heart of Help Me Stream Research Foundation’s premise that reuse is of greater good than recycling, as a portion of that older technology can be repurposed for both niche uses and further research. In the case of the Help Me Stream Research Foundation, those niche uses include repackaging developed-nations’ technologies into “classroom in a box” solutions that are destined for disaster areas and emerging economies that don’t yet have the gift of streaming.
But beyond those niche uses, the Help Me Stream Research Foundation also performs fundamental research on old gear—using best practices gained from partner organizations like Greening of Streaming and its power-attuned members—to prolong the usable life of that gear, which would otherwise be thrown on the recycling heap. As an added incentive, the fact that the Help Me Stream Research Foundation is a 501(c)(3) organization means that we can receive old technologies and offer potential tax exemptions to those companies that want to see their older designs repurposed for use in emerging economies rather than binned.
In other words, I’m arguing that the industry needs to be as innovative with its approach to old technologies as it is with the new innovations we all want to see as part of the lower-power-consumption targets for at-scale streaming delivery.
We’ve seen recent examples of software companies, such as Varnish, optimizing their software to eke more out of older hardware. This is exactly the kind of innovation we need, as emerging economies begin to awaken to streaming’s potential but can’t afford to scale out with the latest and greatest hardware.
The overarching premise of being innovative across the entire streaming lifecycle—from innovative designs that use less power for equivalent performance all the way to the last reuse case for older tech gear—also ties back to a point that Mattera, Kanitkar, and others have made over the past year: The industry needs to solve this problem.
“The way to move forward is for people to know the facts,” said Curwin. “That’s precisely why our initial focus is on producing detailed scientific measurements of energy consumption throughout the supply chain.”
Like Curwin, I’ve had conversations with well-educated consumers who are attuned to sustainability, with each asking how they can help solve the problem. At the U.K. Parliament event, I even presented key findings from a YouGov poll of 2,167 British adults showing that there is marked interest in helping to alleviate the problem from “power streamers” who consume more than 6 hours of streaming per day. But the issues are much bigger than any group of consumers can tackle.
“We think if there were more policymakers, thought leaders, and influencers who can raise these difficult questions and expect deliverable outcomes, it would facilitate far greater impetus for the global streaming sector to find the solutions to make meaningful changes to their own supply chains,” said Curwin.
As part of that movement, we need to not only innovate with new technologies, but to find ways to repurpose the older tech. Only then will the industry best utilize the resources we have at hand while preparing for the next chapter in scaling global streaming to television-sized audiences.
Reducing the carbon impact and energy intensity of streaming media requires an industry-wide focus on both educating consumers and managing their service expectations
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.