Brazil, the World Cup, and the Olympics: Ready or Not?
Extrapolating from the Confederation Cup figures, the data consumption of the World Cup on mobile networks in Brazil is estimated to be around 300TB. Currently a significant TB buyer who is a Limelight or Akamai customer would be paying $0.02 per GB of data transfer. 300TB of data is worth about $6,000 to Akamai or Limelight on this basis.
TIM is a smaller operator than Vivo. They state they expect to shift 11 petabytes per month, so this would indicate traffic of 132PB per year.
If, as the article indicates, TIM has to spend $7.5 million on its caching systems just for TIM’s market share of that 300TB World Cup traffic, then that is a significant amount of money for a small amount of data—particularly given what they state is their existing volume. Surely they already had caching in?
While a different topological proposition, to compare the value of this data to the networks, pricing from a managed service from Akamai or Limelight for 300TB of delivery may be so low that I doubt the cost of flights for the sales rep would balance up.
Something is wrong there.
This compares with the headline figure—attributed to “forecasts from operators based on a survey of Frost & Sullivan's and Cisco“—of 19.2 billion gigabytes or 19,200PB. 19.2PB of CDN distribution would be over $350 million of CDN revenue—and this would be around double Limelight’s global revenue just for their domestic smartphone audience. See the disconnect?
Estimating using the given figures of Vivo’s 77.6 million customers being 65% smartphone users indicates that there may be around (entirely conjectural from me here) 100 million smartphones.
19.2 billion gigabytes / 100 million phones = 192GB of mobile data per user. Not a million miles from the earlier figure, but 192GB per smartphone per year? Really?
My sense is that Wi-Fi data transfer is being included in this, since 192GB would indicate in the region of 30-60 minutes of video streaming on a daily basis. While it is vaguely possible to sit at home and do this, trying to watch a soap opera on the bus during a commute in Brazil would be impossible. The signal variation (from my own experience in Rio and Salvador) is simply too wild to make this an enjoyable enough experience to be a daily activity.
So I am building turrets on castles made from sand here. But you get my point; this is all looking a little weird. The data is clearly wildly conjectural, and the PR teams are clearly giving numbers that tie more into justifying projections for budgets than much that is tangibly measured.
Anyway rather than dig any further holes into numbers that at best I am cautious about, let me make some higher-level comments.
Nowhere in any of these articles is there any discussion of bandwidth. The focus is entirely on data consumption. The network directors who are quoted talk petabytes. The writer himself refers to quantities of images being transferred. The notion is still that mobile networks are purely store and forward. Even the comment about TIM’s 70% reduction in waiting time indicates that the only thinking going on here is “how long will it take my file to transfer.”
And this is where I start to steam at the ears and get a little concerned that the network planning has been done in isolation from reality.
There is a notion that mobile data is used for MMS, or for sending tweets of pictures, for attaching images to emails, or possibly for Facebook photo uploads. More sophisticated users may even record a video and upload that video to Vimeo, YouTube, or any number of other services.
Some users may download these, streaming snack-sized video from their email or social networks. A little buffering, perhaps a pause while the grey bar stretches off to the right, but no real concern if it takes a few minutes for that video clip to get here. Indeed, if it takes too long then the user will just come back look later.
And to be honest I really think that that is about how far mobile operators themselves have looked at rich media delivery on their networks. Just so long as it gets there eventually, and hopefully before the user loses interest then that is just fine.
Well, operators—and this applies worldwide as much as it does in Brazil—you need to pull you head out of the sand. The fact is that bandwidth is going to be the next critical service level you need to focus on. Video conferencing is already starting to drive that. Skype is a killer app for many people, and if not Skype then a myriad of other real-time messaging, voice, and videoconferencing applications that are central to the very essence of mobility.
The reason I carry a mobile is that I want to communicate NOW. All the extra “post”-like data that comes with my email and other asynchronous messaging systems is a nice plus, but to be honest it is feature creep that has taken over the core proposition. What I really want is a clean, fast way to communicate with contacts. Voice, video, IM, I don’t care. Just make sure that always works.
Smartphones just about manage to maintain a traditional phone service, but none of them yet manage to run live video conferencing reliably and well.
And that takes me on to the final part of this article.
The fact that the real service of interest on the mobile network is live video is central to what we should be expecting from Brazil. This is not a chance to let a 100 thousand fans send a photo of themselves in the stadium. Nor is it an opportunity to share clips of key events in the Olympics on personal Facebook pages (until FIFA or the Olympics claim a right to the image and ask for it to be taken down!). No. Brazil should be the Olympics where we move beyond “every event” being available live online, such as the BBC showed the world was possible in 2012. Brazil should be about live video access to every athlete.
And for this we don’t only need planning for PB of data transfer. What we need is huge bandwidth that can handle spikes of concurrency, and with good CDN architecture that ensures that this is well-distributed and load-balanced, and most importantly that the backhaul networks have capacity.
And something else that is simply overlooked in the Globo article, and so perhaps seemingly in all the planning by cellular networks around these events.
During London’s’ 2012 Olympics—as I have highlighted in previous articles— there were many tens, if not a few hundred, cellular multiplexers (cellmuxes) bonding together multiple 3G and 4G links to create high-capacity backhaul links over mobile networks for live video contribution (and rapid forwarding of high-quality rushes) back to TV studios.
Since 2012 the prices of these cellmuxes have dropped significantly. They are being widely adopted by all levels of the TV production industry and webcast industry. They can be found in use on the periphery of many events, being used for deep-action live shots from the crowd, sidelines, front lines, and moving locations. They are providing producers that “must-have” shot from a unique location that was previously inaccessible with a satellite outside broadcast vehicle.
When crews turn up to an event and use cellmuxes they will demand a lot of their local mobile backhaul networks. Their requirement is not a terabyte of data transfer from many thousands of handsets that will wait and try again for ages to post a photo on Facebook. A typical 3G user’s requirement is to transfer a 2MB photo in 10 seconds. Many thousands of those would share 6Mbps and no one would be upset.
Lets factor in some cellmux video production. Imagine we have 100 contribution feeds in each stadium of the half dozen I conjecture may be live at the same time at peak. This could mean that 100 different live video feeds are being sent to international TV stations via the internet from each city. Each of these cellmuxes could demand ~6Mbps.
100x6Mbps equals a further 3,600Mbps or 3Gbps, and this feed is going to be rebroadcast on TV, so if it starves then everyone watching that TV feed will be aware that there wasn’t enough bandwidth.
So where you had already planned, say, a 10Gbps backhaul from your tower thinking that all the users’ photo postings and general use apps, with a bit over for the odd Skype call etc, would be more than happy, what you will find is a small army of cellmux users will soak up all the surplus capacity. Then everyone will start contending for that access. And these linear session-based streams may also congest your peerings too, although these are likely to be hugely overprovisioned.
So I am wondering if it is really likely that the cellular networks will be ready. The data is not, in my mind, revealing enough to say if the mobile internet in Brazil will get “broken” by the World Cup, but I am not entirely confident the provisioning is scoping a wide enough range of use cases. It will be important for Brazil to do some proper planning, not just about a limited subset of data, but about the new emerging technologies that will arrive in Brazil in a few short weeks’ time and begin to pump the Internet full of new content.
In a closed research project, the BBC will deliver the FIFA final and two other matches in ultra-high definition. The project should pave the way for a 4K channel launch in the future.
Multiscreen viewing has finally hit the mass market. Viewers around the globe will be able to view every angle of play on any connected device.