IBC Q&A: Akamai’s Bishop on Rio 2016 streaming, future of live 4K, VR streaming
Akamai Technologies arrived at IBC 2016 fresh off a historic effort at Rio 2016, where NBC Olympics leveraged the company’s content-delivery network (CDN) and technology to support its monumental streaming operation, which comprises 3.3 billion total streaming minutes, 2.71 billion of them live. Also, 100 million unique users tuned in to NBC Olympics’ digital coverage over the course of the Games.
At the show, Akamai showcased a variety of its offerings, including Ingest Acceleration, Dynamic Encode to Entry-point Mapping (DEEM), Media Acceleration (including a demo of sub-10-second low-latency 4K-video delivery over the public internet), Ingest Monitoring and Reporting capabilities, and the production model for its Broadcast Operations Control Center (BOCC), which launched earlier this year.
SVG sat down with John Bishop, CTO, media, to discuss Akamai’s historic effort with NBC Olympics in Rio, how the BOCC operation is progressing, the state of live streaming in 4K and VR, and how he sees video-streaming data-rate capabilities increasing in the year ahead.
What insights did Akamai gain from the Rio Olympics experience in terms of overall Internet traffic, and how do you use it evolving for future live coverage of high-profile sports events?
The hard thing about seminal tent-pole events like the Euros or Rio is that the internet still goes on with the same amount of traffic. One of the most amazing things about Rio is that the traffic numbers were huge but the internet didn’t hiccup along the way. Our BOCC monitored the whole thing, and, from the moment the Opening Ceremony started, the audience size compared to London or Sochi began to increase by orders of magnitude in terms of velocity, overall traffic, higher data rates. The connected devices and set-top boxes numbers were through the roof. At that point, you knew that this was going to be a historically big streaming event.
The Opening Ceremony set the stage for the peaks ahead: Bolt, Phelps, Ledecky, women’s gymnastics. But those events were like big cardiogram spikes. Yes, we definitely saw a surge in audience, but it wasn’t 0 to 100 [mph]; it was more like 60 to 85. But we already did that 0-60 sprint, and the audiences remained spread across all these events, including at times 70 concurrent events.
From a U.S. perspective, it was unclear how this event was going to play out, because the Rio time zone was so similar to the East Coast. In many ways, you’re competing against television, because you can go to your cable service and watch those same events at the same time. But one of the most interesting things was, we didn’t see the audience dipping significantly during primetime. People just kept on chugging right through.
One of the biggest surprise for me during the big [viewership blips] for big events midday and mid-workweek days were how many people were using connected devices. We absolutely expected [viewership increases] on tablets and computers, but we saw a massive amount of people watching through Apple TVs and Rokus. Of course, we had more connected devices [streaming] in the evening, but it was a nice surprise to see connected devices stay strong during the day.
At NAB 2016, Akamai launched its BOCC. Can you provide an update on how it is progressing?
Over the last 15 years, the one goal that we’ve had for internet video is always for the experience to be television-like. But, today, we’re not trying to be broadcast-like anymore; we’re trying to be broadcast-equal or better than broadcast. I think, in many ways, we’ve gotten there. I think Rio and the Euro Cup and some things we’ve done since show that we’ve passed broadcast in terms of fidelity. We’re doing higher resolutions; we’re doing things like 4K and VR and 360 experiences — things that are impossible to very difficult to find over traditional TV services.
We are nearing [broadcast level] in terms of concurrency, but one area we are really trying put that same level of rigor into is hygiene and telemetry [of the live stream]. Broadcast systems have a broadcast operations center with detailed telemetry and instrumentation. If we are really trying to replace television, we need to have that same level of hygiene and make sure that every system component is being watched in terms of health, load, and compliancy. We need to be looking at every part of the supply chain. We want to be able to say we have the same level of rigor in OTT that we do in broadcast.
How is Akamai working to decrease the latency seen in live-streamed sports events versus linear television?
On an average [live production], you’re probably 45 seconds behind live. And nothing annoys me more than when I get an alert on my phone that somebody just scored and I haven’t even seen it yet. It’s just been the way that internet video has always worked. At NAB , we put a stake in the ground and said I think we can get that 45 seconds to sub-20 seconds. Here at IBC, we’re showing in the 12- to 16-seconds range for low latency. [Given] that, by and large, most of the internet streams today are at 60 seconds of latency, that’s an 80% reduction. In the U.S. market, most cable latency is around 6½ seconds to 8 seconds, so it’s becoming a pretty parallel experience. I think, in the next couple of years, we’ll continue to shave that, and I can see us getting sub-10 in the near future. But we want to be careful [that,] for our premium media customers, we don’t provide ultra-low latency at the expense of quality. We’ve seen that, if that buffer wheel comes up, people leave. There’s low to no tolerance for that.
What do you project for the rest of 2016 in live-streaming sports?
First off, we were extremely happy with the Euros and Rio  in terms of the concurrency [of viewership]. We were starting to get to event concurrency that approaches what Game of Thrones would do on a premiere night. We’ve got line of sight to having the biggest global audiences. Could the internet support 113 million people on the Super Bowl tomorrow? That would be a challenge to deliver a reliable, compelling experience. But that’s what we’ve been building towards.
In terms of [the rest of 2016], we’ve already entered college-football season, NFL’s already started, and the blip [from the Olympics] is becoming the new norm. I’m very happy with the way college football and fall sports have started off, not attributed to any specific event. We’re seeing double-digit year-over-year growth.
Have you seen growth in the live streaming of sports in 4K over the past year?
There we’re seeing two, three, four maybe a month like a big soccer tournament’s final match. We did some stuff around 4K with the Olympics as well. But compare that to where 4K used to be not long ago when it was one [event] per quarter; that number is now several a month. I think we’re beginning to see that with VR as well.
Last spring, we did one of the first 4K events with a 15-Mbps top end, and 83% of the audience that watched for 2.5 hours came attached at 15 Mbps. The average connect rate was 13.1, which is very high. So, if I’ve got an audience that can attach at 13.1 Mbps in April 2016, I’m looking forward to 17 Mbps. All those numbers are just going up, so I’m not concerned about UHD from a distribution perspective at all. I think we’re ready for that 15-Mbps sweet spot.
Do you see the number of sports events live-streamed in VR increasing, and how can the live-streaming VR experience improve?
There are a lot of distribution methodologies, and we’re involved in all of those in an effort to establish early best practices. Off the top of my head, I know of five VR events that will happen by the end of October. Compare that to one or two every six months. By NAB and IBC next year, we will be talking about 50-plus VR events that we have [streamed] successfully, which is a big deal.
There is not a lot of good VR. A lot of the VR today is [streamed at 8-10 Mbps. The VR experience should be crisper because the whole point of VR is that it is immersive, so it’s got to be lifelike. I think we’re going to get there, but I think it’s going to take a lot more bits. If HD today is at 6 and Ultra HD is at 15, the is it 30 or 40 for truly good VR? There’s going to be wild experimentation with what’s the right data rate and the right codec. Are we H.264, or H.265? There are at least five techniques right now on VR that are all slightly different. Some of them are trying to optimize for quality, some of them are trying to optimize for bandwidth, some are trying to optimize for interactivity or even monetization. That’s what’s going to vet out over the next 12 months. But I think there’s an appetite for it. It’s not 3D, that’s for sure.
Do you believe streaming rates will continue to get higher as viewers demand better and higher resolution experiences?
Yes, definitely. When I look back to like last year, it was very common in an ABR [adaptive bit rate] ladder to see the higher data rate around 4.5 MBps. Right now it’s very common to see north of 7 in the ABR ladder. An average connected device is 5.5 to 6 MBps or higher, which wasn’t even available last year. So I think that’s driving the next experience.
When you look at what we saw with Rio or the MTV VMA’s, where the traditional audience was on the decline and the internet audience was exploding, there is going continue to be massive growth in internet traffic.