Your browser is not supported. Please update it.

26 June 2014

So how much bandwidth should telcos and cablecos plan for

There has to be a huge dollop of doubt thrown on any research that is commissioned by a cluster of vested interests in broadband, and conducted by their partners, relying on data supplied by things like the Cisco VNI, itself a compilation of 8 or 9 other sources of data.

The danger is twofold, that it could end up as an average derived from a bunch of averages, and that it can be made to serve the purposes and reach the right conclusions for the sponsor.

And we feel that all of this is present in the piece of research put out this week by a Dutch consulting group Dialogic, working with Eindhoven University of Technology and commissioned by NLKabel and Cable Europe.

NLKabel is made up of the Dutch Cable groups, and Cable Europe is the lobby group for cable that runs across Europe. Both are dominated by the towering market presence of Liberty Global and if this had been the dictated wish list of Liberty Global CEO Mike Fries, the results would not have read any differently.

The conclusion that we are asked to draw is that European households will need 40% more downstream internet data each year through 2020, reaching a point where 165 Mbps will be required at an average household, and that 20 Mbps would be needed in the upload.

Or to put it another way, the current and next generation of broadband technologies for cable can cope easily, and the Telcos will be hard put to keep up, even if they get to G.Fast before 2020. All of which is the precise sweet spot that Liberty Global might have asked for if it was simply dictating the results.

However the methodology is complex, though by no means thorough, and it is worth taking in some of the finer points, since there is plenty of truth here in between the suspect data.

A simple rule of thumb might be that most households would surely top out at a number of concurrent 4K video streams into the home, delivered in 15 Mbps Adaptive Bit Rates streams, encoded in HEVC, which in homes with under 2.0 children per home (the Dutch average is 1.7) should be no more than 4 of these streams, so around 60 Mbps.

So how can this report justify 2.5 times that bandwidth? Well there are issues of inefficiency, and overhead, there is the internet of things and how it will be spurred on by a bewildering number of internet connected devices in each home, and one of these will be permanent HD video feeds for the purposes of security. Or will it?

This is the problem with this type of forecast, in that you can make a mistake really easily with any set of assumptions. If every home also has a broadcast TV service in it, as it does today in the Netherlands, where this research was done, then there is an alternative route rather than the internet right now, to bring content, both 4K and full HD, into the home, in bulk and to store it.

Any internet traffic has to contend with this incumbent content for access to the home network, which is invariably WiFi in the Netherlands or worse still, uses a powerline backbone to move data around the home. You won’t currently get 4 x 4K video from the internet and deliver it around the home, so what would be the point of demanding that kind of bandwidth?

Let’s go back to that issue of security video. Do people really want to have a camera on them all the time? This is highly unlikely in the freedom obsessed Netherlands. Security cameras typically use lower resolution than HD, and frame rates like 5 frames per second, to keep down the video bandwidth.

So let’s look at this as a process. If you install home security right now, with existing monthly data caps, you will probably not want to have those video cameras on full time of you will pay more. And if you did they would have a two way effect. This type of video drives uploading, while wherever you are receiving it, this will add to the download traffic.

So during a period when uploading is slow, surely you would put a motion sensor in place, and ideally try to store the video locally. That way the video in not on very often, and when it is, none of it has to leave the house, unless the owner is worried enough to view the video on his or her remote smart phone.

We admit that most home automation security services export this video to the cloud, but that may change as cheaper storage becomes available, or cheaper processing. In which case it may be motion triggered, and then analyzed by powerful servers in the cloud to establish if that movement is a burglar or just your cat coming in through the catflap.

If such a service is set up in that way when bandwidth is tight, then even if tons of bandwidth becomes available later, there is no real reason to set it up differently. So we’re not sure where such increases in bandwidth requirement will come from for video security.

The report is littered with these types of concerns. Another is that the HEVC codec is not mentioned once in the report and this alone would halve all video data, and neither is it discussed in an alternative form such as H.265 and there is no reference to increasing processing power that would lead to even faster codecs in the forecast timeframe, creating less of a strain on bandwidth.

The report concludes that a 40% increase in CAGR will occur for downstream bandwidth requirement and a 44% surge in upstream traffic demand. It also says that there will be power users with a a requirement over 1Gbps in 2020 (see diagram below) and an upload need of 314 Mbps. This should only be expected from 2% of broadband customers, the report says.

The methodology has been to take each type of traffic and see how it has grown recently and then work out how many more people will be doing this activity, and how much each user will grow and project the forecast. But it also adds some mysterious ‘as yet undiscovered’ revolutionary services, as well as a separate growth per user being added to this and reconciled against it. It is an approach that is bound to lead to a high bandwidth requirement.

The limitations of in-home WiFi networks have not been taken into account, which would prevent very high throughputs into the home being helpful and the disincentive of existing caps and the fact that if you have plenty of video streaming services, then it is less likely that your P2P piracy would continue going up ‘ these seem to be mutually exclusive, and the assumption in this report is that more people would engage in piracy and do it with denser video, and we’re not sure that is sustainable alongside increasing streaming video.

This is especially true given that one source of data came from equipment provider Sandvine which monitors current internet traffic. It has suddenly noticed that in Europe from a standing start in 2013, real time entertainment now takes up 17% of the bandwidth it monitored. Well if it only has one reading, it’s a little early to suggest that this will grow and grow, without any negative effect on piracy.

As iTunes grew the resulting music piracy tailed off, so history tells us that this is likely to happen with video.

The graph below shows the forecast for bandwidth requirement.
While the upload speeds show the required bandwidth broken out by application.

There seems to be a rising overhead built into the calculation, which includes packet acknowledgements as well as the overheads from putting data into packets. As far as we can see the internet usage was measured and yet none of the measured traffic was assumed to be overhead, so it looks like some double counting has gone on here. Overhead appears to come out as 10% of traffic.

Another trick to the report is that the university interviewed 10 experts, and took what they said as gospel for adding services. If most of these were from cable companies, then it is not surprising that they assume that there will be a surge in the number of connected devices and accompanying services in a household, including the introduction of advanced, surveillance applications and that home automation will also play its part. We have criticized these ideas on multiple occasions in that tiny amounts of reporting data used in IoT apps, sent once a month, cannot add up to the type of bandwidth flow that people are suggesting for the Internet of Things, with the exception of video for security and we have already pointed to how that has been exaggerated.

The other suggestion that came from experts which has been added is the take-up of distributed computing, in which personal devices solve computation problems, such as SETI@Home as well as the widespread adoption of a digital currency (e.g. Bitcoin). Neither of these things are a given and because cost per transaction will be very important in the uptake of online currencies like Bitcoin, they will have to be efficient.

Given that portable devices are already offloading close to 80% of their video and data activities to fixed line networks, via WiFi, it seems odd that there was no calculation carried out on a type of device basis and forecast forwards for offload growth.

The main categories of usage for existing applications were web browsing, email, social media, remote backup, conversational apps, online video, remote workplace, online music, file downloads, online gaming, P2P file sharing and personal cloud storage

Sandvine calculates that the average online video usage is already 217.5 Mbytes a day, which let’s face it, is not even 30 minutes of Netflix programming, so we can see that this will have to rise substantially, as OTT SVOD services begin to be taken up across Europe.

But much of this bandwidth may come off the piracy figures, and we really think that forecast should be left to analysts, not sponsored by vested interests, and then they would be more believable.