Is This the End of the CDN As We Know It?
Recent news signals to many the imminent death of traditional content delivery networks (CDNs). But does it really?
When it comes to streaming, there are often compelling issues with the two sides to the content coin: creation and delivery. On one hand, the goal many of us shot for in the late-1990s as streaming came into its own—the democratization of content creation—has become almost second nature, what with the phones in our pockets and cameras all around us.
On the other hand, the democratization of delivering content didn’t really emerge the way most of us had planned. Apart from one-to-one delivery (aka, FaceTime or WhatsApp video calls) most content today is delivered through some form of content delivery network (CDN).
Defaulting to HTTP
What started as a way of accelerating the delivery of file-based assets (from video games and operating system updates to on-demand theatrical “streaming” content and pre-loaded theatrical trailers) gave way to the much more difficult task of streaming live content to the masses. But the task of streaming at low latencies was too big a hurdle—both from a technical standpoint, since low-latency streaming had the heavy lift of unicast streams with its required ratio of one stream for each consumer, as well as from a cost standpoint, from the heavy-iron servers and costly streaming server software. As a result, the CDN market dropped back and punted towards plain-vanilla (HTTP-based) video delivery.
That HTTP approach was a hybrid mix of live streaming (albeit often delayed 30-or-more seconds from the origin server) and small-file playback (the segments or chunks or fragments being delivered to a consumer’s device and played back in order, as long as they arrived in a timely manner). And while it brought the cost down in the short term, it hobbled all streaming efforts when it came to “live television” streaming to the masses.
On the surface, though, the evidence for several years appeared to contrast the doomsday scenarios that the naysayers brought to the argument. As far back as 2011, with the ratification of Dynamic Adaptive Streaming via HTTP (DASH) by the Motion Picture Experts Group, a number of red flags went up warning of a potential “painting ourselves into a delayed delivery corner” with the typical 30-second delays that HTTP-based delivery entailed.
These warnings were raised in Streaming Media magazine, on online forums, at StreamingMedia shows, and at other industry events. And, as part of full disclosure, I have been one of those naysayers for at least fifteen years, even while arguing in defense of the fragmented MP4 (fMP4) approach in a 2011 joint white paper for Adobe and Microsoft titled Unifying Global Video Strategies.
The mainstream thinking around HTTP delivery, however, prevailed, with many off-the-record conversations going something like this: It’s true that we’re painting ourselves into a corner when it comes to delivering real-time content at scale, but HTTP-based delivery lets us gain market share quickly—at the expense of cable and over-the-air (OTA) broadcasters—and then eventually we’ll see whether consumers actually want live delivery. But we’re betting on the fact that we don’t think they really want it.
Enter LL-HLS
It turns out, though, that they really did want it. Whether for sports (or sportsbook) or for bi-directional delivery for distance education and live events, the need to return to low-latency delivery approaches returned to the fore over the last half-decade of streaming delivery.
The first cracks in that thinking came, ironically, with the advent of a low-latency streaming specification in 2019 meant to address the 800-pound gorilla in the room: Apple’s HTTP Live Streaming (HLS).
At that point, Roger Pantos was pushing the Low Latency HLS (LL-HLS) specification through the use of what we’ll call advanced HTTP. The IETF spec was presented at the Streaming Media West show in late 2019, with suggestions around fMP4 and HTTP/2 push, just in time for the pandemic to upend and hypercharge streaming delivery.
Health of the CDN Industry
Since then, CDN usage has continued to grow, with Mark de Jong, founder and chair of the CDN Alliance, noting last year that “72 percent of Internet content is delivered through CDNs” and likening the effect of a CDN to your local electrical company.
And he’s not the only one to point this out: Bloomberg’s Nate Lanxon in 2021 made a similar point about a widespread outage from Fastly: “You don’t hear much about content-delivery networks, or CDNs, until they stop delivering. A global outage of major websites on June 8 that lasted about an hour was caused by problems at the San Francisco-based company.”
When I asked de Jong about the health of the industry, in an interview just before press time, he didn’t mince words. “It’s not healthy, as was made clear by a number of companies scaling back their CDN services or [shutting down] completely,” says de Jong, using recent departures Edgio, Lumen, and Stackpath as examples. “And I don’t think we’re done yet, as I expect at least one more to go down this year.”
But de Jong argues this is better for the overall health of the industry. “Clearly there were too many players, and too much business had been bought,” says de Jong. “CDN delivery has been a bleeder for a lot of companies, so they tried to make up for that with other services and companies sometime might have bought deals to make sure they got the deals in.”
Like the electricity industry, however, he argues that the bad news has been a hiccup, not a death knell. “Both of these industries have companies involved are still almost invisible—until something goes wrong,” says de Jong. “But now we are getting to a point where pricing is at what may be a sustainable level. Think of the energy companies: their capacity is needed but the margins are low. If you’re good at what you do, you’ll survive.”
With this acknowledgement that pure CDN delivery has become a commodity, with the exception perhaps of several geographies and use cases (war rooms, sports, etc.), the question then arises as to whether the days of CDNs are numbered. Mark de Jong’s response is an unequivocal "no."
“With certain content creators, they’re big enough to strike deals directly with ISPs,” he says. “But basically everyone, with exception of Netflix, still relies on third-party CDNs to some extent. Yes, it’s become easier to build your own CDN, so there will continue to be a proportional shift in terms of reliance on CDNs, but it’s not a major shift.”
Finally on my question of kicking the can down the road, when it comes to HTTP-based delivery, de Jong argued that advances in HTTP may yet yield the possibility of near real-time delivery using a derivative of HTTP known as the recursive QUIC (like GNU meaning GNU’s Not Unix, QUIC initially stood for Quick UDP Internet Connections) that’s become somewhat synonymous with HTTP/3.
“Latency is both protocol and infrastructure,” says de Jong. “Yes, you can spin up instances of third-party clouds, but you have to tweak the underlying infrastructure. And there are limitations with certain protocols, such as WebRTC. Only Edgio fired up WebRTC at scale; none of the other CDNs followed suit.”
But if Media over QUIC (MoQ) takes off, de Jong contends, “it’s going to drive us back to HTTP at low latencies and regular CDNs can run it at scale. For a lot of use cases, MoQ may be a better approach than to use WebRTC.”
The CDN Alliance will be discussing this and more during its first virtual roundtable, set to be held on January 28, 2025 at 5pm CET. The session is currently sold out, which is a good indication that interest in CDNs remains strong.
Related Articles
Media Over QUIC aspires to streamline the entire streaming process by bringing both contribution and distribution into a single protocol again, reducing the need for intermediary transformations. Since MoQ is still in the early stages of development, it will take some time to see it deliver on this promise.
05 Dec 2024
Why are we still talking about the challenges of low-latency streaming? Put another way, with all the hype about real-time streaming and its evident advantages for high-interactivity applications ripe for monetization like iGaming and sports betting, why haven't we made more progress toward realizing real-time and its potential? Dolby's Ryan Jespersen, AWS's Evan Statton, and 2G Digital's Allan McLennan discuss in this clip from Streaming Media Connect 2023.
13 Dec 2023
Eric Carson, Chief Revenue Officer of Ateliere, writes about the importance of democratizing both content distribution and production around the world, along with the ways that Ateliere technology is helping to make content both widely available and accessible.
12 Dec 2022
MPEG DASH is the latest hot topic in the online video space. Here we break down what it is, and what its implications might be for video delivery in the future.
22 Nov 2011
Understanding this hybrid delivery approach is key to advancing robust streaming solutions.
09 Feb 2010