Cord cutting is all over the press. Disney+, AppleTV+, HBO Max, Peacock, and many other services are launching to fill the gap. There is an absolute flood of content and a bevy of service providers poised to go over the top, directly to consumers. But what about the OTT plumbing? How is this flood of content physically going to make its way to the end users?
The most obvious answer is, “Don’t worry, the CDNs can handle it.” Well, not so fast. To keep the water metaphor going, let’s dive a little deeper. Cisco’s Visual Networking Index (VNI) forecasts 396 Exabytes of IP traffic per month by 2022. 325 Exabytes of that, according to Cisco, will be video. How much can the top tier CDNs handle? Well currently, tier 1 CDNs have roughly 400 Tbps of peering capacity. Just noodle around on their websites and in their annual reports and you get to a number pretty close to that. If you ran the tier 1 CDNs 24×7 at full capacity for a month, they can deliver roughly 130 Exabytes, assuming steady global traffic with all of the capacity perfectly matched geographically for demand. We know that traffic varies widely across geographies based on time of day, so it is safe to assume that CDNs realistically can only deliver at best 1/2 of their advertised capacity over an extended period – so roughly 65 Exabytes per month. That is only 1/5th of the forecasted demand for video. That’s right. We are going to need 5 times the capacity, roughly 2 Pbps, to handle the forecasted video demand, in just a couple of years. The capacity gap is huge.
The Looming Content Delivery Capacity Gap
Well, maybe CDNs can just grow. But the competitive nature of CDNs has already driven most of the profits out of the content delivery game. That is why CDNs all seem to be turning to security services as the new profit centers as they search for better margins. Ask any CDN salesperson and they will confirm. It just isn’t plausible to expect tier 1 CDNs to build 400% more delivery infrastructure in 2 years to accommodate growth in a low profit-margin video delivery service. Even if they did, the ISPs couldn’t handle that kind of traffic increase. For a CDN to grow, they need to increase their infrastructure (servers, racks, power, cooling, electricity) and their peering arrangements. Delivering more video streams means creating the streams and then passing them off to the ISPs for delivery.
But the ISPs have little interest in massively increasing their peering capacity. If they receive more streams at their peering points, they have to carry those streams throughout their network, absolutely flooding their core. In fact, Akamai itself did a study of ISP capacity in 2018. They concluded that ISPs have more than 70 times more bandwidth in the last mile than in their core. In other words, for every 70 Mbps they can deliver to an end user, they can only support 1 Mbps in their core. Why is that true? Well ISPs’ networks were built with the expectation that only a fraction of their subscribers would be downloading something concurrently. In the days of wired telephones, phone companies could only support 5 concurrent phone calls per 100 phone lines for the same reason. You probably never noticed it, unless you tried to make a phone call right after Rolling Stones tickets went on sale. For ISPs, on Saturday night at 9:00, that whole concurrency assumption goes down the drain. Imagine if everyone in the city decided to take a shower at the same time … the plumbing just can’t handle it.
Conceivably ISPs could massively increase their core networks to alleviate the flooding, but they don’t have a financial incentive to do so. ISP contracts with their customers are typically based on peak bandwidth, not usage. More usage by end users doesn’t translate into any additional revenue for the ISP. Why then, would they agree to massively spend on infrastructure?
The punchline here is that OTT demand is starting to overwhelm capacity, and there is no financial incentive to grow the capacity – either for the CDN or the ISP. Furthermore, OTT demand is about to absolutely explode, just as the bottlenecks are starting to fill up.
In Qwilt’s view, there is a very straightforward solution to overcoming a bottleneck … avoid it. The only part of the ISPs’ networks that can handle the coming traffic demand is the last mile, so just move the stream creation there, at the ISP network edge. All it takes is some edge compute, some storage, some software, and a coordinating entity. You get something that looks and functions like a CDN, but its core elements are owned by ISPs who have the needed financial and operational strength, as well as the incentive to collaborate with content publishers. They also have racks, power, etc. in the optimal locations, and engineers on site to provide support.
This is the thought process underlying Qwilt’s Content Delivery Sharing service. It works a lot like Uber. In this case, local ISPs are the drivers and the requested video streams are the riders. The principles of the sharing economy fit exceedingly well here. Instead of a single entity owning all the infrastructure centrally, it is owned locally, by the ISPs, those providing the actual delivery services to the end users. The ISPs can build edge nodes with off-the-shelf Linux servers and Qwilt software. Then they place the nodes in the last mile. Qwilt works with content providers to get them on board with having their content delivered this new way. Then, when an end user selects a piece of content to watch, their request is automatically directed to their own ISP to find the content nearby. It is fast, low-cost, modern. Best of all, it bypasses the bottlenecks.
Qwilt Content Delivery Sharing
Content providers, ISPs, and end users all benefit simultaneously. The content providers get the delivery capacity they need to be successful; the ISPs keep their customers happy without having to massively grow their core; and the end users get tremendous quality-of-service only possible when content is stored right near them. Qwilt is working with some of the largest ISPs and content providers in the world. The underlying technology is referred to as “open caching” and is endorsed by the Streaming Video Alliance, so it is widely accepted. Although they may not initially see it this way, the CDNs actually benefit as well. They get to focus on higher margin services like security without having to massively increase their capacity to satisfy their customers. Content Delivery Sharing is a win for everybody.
I talked about Content Delivery Sharing at the recent Intel Visual Cloud Conference at IBC, and highlighted a case study with Globo. This media giant has partnered with Qwilt to expand content delivery and improve QoE for viewers across Brazil. Take a look >>
Intel Visual Cloud Conference @ IBC 2019 (Jeff Gilbert @ 3:10)