Updated One of the toughest jobs in this industry is analyzing out the latest Cisco Visual Networking Index. First you have to get your head around all those numbers which are impossible to visualise, like exabytes. And then you start to see questions in the logic, and wonder if perhaps Cisco has gone too far off in one direction or another.
The release this month has a great start, reminding us that by 2021, more people will be using mobile phones (5.5 billion) than have bank accounts (5.4 billion), or running water (5.3 billion), or landlines (2.9 billion).
Oops, there’s the first one. There's nothing like 2.9 billion people with access to fixed line phones. More like 1.15 billion – at least that’s what we have always been told. And we have done a count of major operators. They reached a peak at around 1.3 billion and since then they have gone down not up? So if all the Cisco numbers are as far out as that, then it makes you wonder why we have to read any further?
The next thing that really creates disquiet is the blithe way that everyone assumes that 5G is going to be here in 2020, even though it is likely that the standards won’t have been with us long enough for devices to have been built.
This is Cisco, a vendor, believing what other vendors, and to a lesser extent, operators say. There will be advanced networks and some of them will use some elements of 5G by then, such as New Radio.
But 5G has an inherent problem, in that it is being designed to reach multiple, simultaneous objectives – low latency, high video speeds, an order of magnitude improvement in the number of connections. We think suggesting that by 2020 over one per cent of cellular data will go over 5G, is like making supporting noises for all those ambitious operators out there. Especially since AT&T and Verizon, two of the most ambitious there are, are going with fixed broadband first in 2020, because then it will be easy to leap ahead of the standards. Will that count as cellular? We think not.
The truth is that there are two major things going on creating all these changes. People are moving from 3G to 4G, and a lot of connections are still not 4G, or are not even on a smartphone, and it is these that define the data increases.
Cisco correctly maps the arrival of new services which eat data, like Facebook or Snapchat video, but we’re video experts, and video can get smaller as well as larger. Many OTT video services are particularly wasteful of spectrum, and to some extent it is that which is driving the requirement for 5G speeds.
They don’t use the latest codecs, and their ABR (adaptive bit rate) set ups are designed to discover LTE and drive up data usage fourfold. This is partly because in order to get people off 3G operators have to lure them with big data caps, but also because if a video takes 2 Mbps in 3G, it will take 8 Mbps in LTE, so if people do not have bigger caps in LTE, they would be able to watch even less video while paying even more money.
Codecs are on the march to improve that. HEVC is now passe, even if it hasn’t sorted out its royalty pool, and codecs from AO-Media and others are threatening to drive bandwidth usage down a notch, maybe 30 per cent next year, and further the year after. Before you know it UHD video will take the same bandwidth as HD, and there won’t be the same data increase. And anyway phones for the most part do not have UHD screens, so you could get increasing use of video and decreasing data – it’s possible. And anyway, most of this will continue to be carried over Wi-Fi, until cellular offers all you can eat data or is really, really cheap.
It has to be agreed that this is happening in the US, but that all you can eat data mentality has yet to catch on internationally and the US won’t drive this data on its own – not with a mere 330 million smartphones by then.
And that’s the other thing, getting smartphones to the rest of the world, it’s non-trivial. Ask an operator in Africa if he wants everyone to have iPhones? The networks would break if they did, and since they only get them second hand, and two generations late, it will take a lot longer than Cisco is suggesting.
Bar chart displaying compound annual growth rate (CAGR) of various device types from 2016 to 2021.
Here is the Cisco take on what happens to phones – notice that non-smartphones go from 41 per cent of the global market to just 13 per cent in the next five years.
We are not saying that people in those markets would not want these phones, but they have to work with the network, and even then the networks are not provisioned to deliver that amount of data and at current rates the consumers could not afford it.
Operators always just keep driving data usage, because they want more money from their customers. If they cannot get more money or if they have to give it to them for the same amount of money, they will drag their heels.
]]One of the core drivers cited by Cisco is Virtual Reality where people shove their smart phone into a headset and walk around even less aware of physical reality than they are today. It’s fine for a simple demo, but VR on phones right now is back where video was in 2000, it is just something to show once to someone for them to “coo” over, but not to use on a daily basis.
And we accept that in developed markets VR will emerge experimentally in the apps of the big social media players, and will get used at scale quite quickly. But not for large periods of the day. If that happened, you could expect a 10x increase in data overnight. It will be slower than that.
No one has yet worked out “how” to send VR anyway, and the best effort sends one screen full of data at full resolution, and 9 overlapping tiles in each direction at far lower resolutions. Okay on 5G the latency may be good enough to switch tiles in 100 milliseconds, but it’s not right now and that’s what we are working with. So data speeds now are simply double, and the Quality of Experience is appalling.
Cisco lists these milestones by 2020, there will be 1.5 devices per capita, nearly 12 billion mobile-connected devices including M2M modules. Mobile network connection speeds will increase from 6.8 Mbps in 2016 to 20.4 Mbps by 2021. That’s very reasonable.
Machine-to-machine (M2M) connections will represent 29 per cent (3.3 billion) of total mobile connections — up from 5 per cent (780 million) in 2016. Now that is a serious reduction in the rate of IoT instances from early forecasts which toyed with 25 billion to 50 billion in this timeframe.
Cisco says 4G will support 58 per cent of total mobile connections by 2021—up from 26 percent in 2016 and will account for 79 per cent of total mobile data traffic. Again, all reasonable.
It is the next stat, that the total number of smartphones (including phablets) will be over 50 per cent of global devices and connections (6.2 billion) — up from 3.6 billion in 2016, that drives all the traffic. But we find it hard to believe. And connecting the last 2 billion is likely to slow down the average amount of data used, not increase it.
Cisco says by 2021, global mobile data traffic will reach 49 exabytes per month or 587 exabytes annually) and that live video will increase 8.7 times from 2016 to 2021. By live video we assume this is sending video person to person, as opposed to viewing live events (see elsewhere in this issue) because that is already falling if the Superbowl is anything to go on. Mobile video will represent 78 per cent of all mobile traffic by 2021.
Cisco forecasts the rise in VR headsets going from 18 million now to 100 million by 2021 — a fivefold increase. We have a serious concern with this. It’s not something you can do in a room without injuring yourself, you cannot do it walking down a street, and it cuts off social interaction – it’s a fad, little more and there is no evidence it will expand to the size of the market.
We think that the experience will have to get a lot better and a new form factor will need to be invented before that happens (although that will happen).
Cisco thinks that VR traffic will grow 13.3 petabytes a month, to 140 petabytes a month in 2021. And that’s reasonable, but we’re not sure about the headset numbers.
Finally, the idea that the connected wearable devices market will drive M2M just loses us. Wearables have fallen off a cliff. Cisco analysts have drunk the Koolaid.
Coming to an area that we understand, Cisco says that 60 per cent of mobile data was offloaded and that this will only rise to 63 per cent by 2021.
Just a shift in screen resolution alone and the introduction of more dense video feeds, will make that estimate too low, unless operators throw in enough data into packages to watch video for 3 or 4 hours a month.
Cisco also says that public WiFi hotspots (including homespots) will grow sixfold from 2016 (94.0 million) to 2021 (541.6 million) and that Wi-Fi traffic from both mobile devices and Wi-Fi-only devices together will account for almost half (49 per cent) of total IP traffic by 2020, up from 42 per cent in 2015. Now that’s a number we can agree wholeheartedly with. Long live Wi-Fi.
Copyright © 2016, Faultline
Faultline is published by Rethink Research, a London-based publishing and consulting firm. This weekly newsletter is an assessment of the impact of the week's events in the world of digital media. Faultline is where media meets technology. Subscription details here.