Chapter 3
Can Demand Grow Indefinitely?
Volume Versus Speed
Demand for mobile services is typically measured in monthly
data volumes—a metric of the quantity consumed. However,
the discussion so far has suggested that the key reason for
introducing new generations of mobile technology is speed.
This chapter discusses how capacity and speed are related
before considering the likely growth in demand for both speed
and data in 5G deployment.
A simplistic approach might assume that there is no
relationship between speed and data. For someone who wishes
to download an attachment or a video, the volume of data is
unchanged regardless of speed—it just takes longer on a slower
connection. In practice, there is some correlation. Users will
only attempt certain activities, such as streaming video, when
the data rate is high enough. Hence, once they pass certain
thresholds, higher data rates tend to trigger step-changes in
demand. MNOs have observed this effect by noticing that data
demand tends to rise significantly as subscribers who have
moved from 3G to 4G get faster and more reliable data.
However, it might be expected that, once speeds are availablebeyond those needed for data-intensive applications, the link
between speed and demand would again fade away.
New generations of mobile technology typically aim both to:
1. Provide higher data rates in order to improve the user
experience.
2. Provide more efficiency in terms of throughput per unit of
radio spectrum to allow networks to carry increased traffic.
This chapter considers the extent to which this is needed in 5G.
What Speed Is Needed
The data rate that is considered acceptable has progressively
grown over time. As higher speeds have become possible,
application providers have delivered new services, driving the
demand for these types of speeds. For example, once speeds
rose above 1 Mbps, delivery of video content became possible,
spurring a huge growth in demand. This then led to the delivery
of higher resolution video services, which again spurred
demand.
Determining how much speed is “enough” is problematic.
The speed that is sufficient at a given point in time for the
services typically consumed can be calculated. New services
may be devised that result in a greater demand for speed,
although this is tempered by the fact that apps are increasingly
being developed for a global market. China and India are far
larger markets than Europe but have lower speeds of
connectivity. Hence, even if very high speeds are provided indeveloped countries, the focus of app developers may
nonetheless be on lower speeds to access larger markets
(techUK 2016).
Demand for the highest speeds and the largest data volumes
is almost invariably driven by video consumption. A person can
only watch one video stream at a time, so understanding the
data rates associated with the highest quality video feed
required is a good upper limit. This may increase in the future
if higher resolution video grows (e.g., demand for 4K video) or
if applications such as virtual reality demand more
information. By way of example, 4K video (video with roughly
4,000 pixels of horizontal resolution) requires around 20 Mbps.°
This is an upper limit, as most mobile screens are far too small
to make watching video at this resolution worthwhile. MNOs
have found that “throttling” video back to 1 Mbps or even less
has no noticeable impact on mobile handset users.
A somewhat different question is the speed needed for
instantaneous web browsing. The issue here is less about
absolute speed and more about “latency”—the time it takes for
a request (e.g., for a new page) to be sent to a server and to
receive a response (as shown in Figure 3.1). Beyond a certain
speed, other factors—such as the turnaround time at the server
and the delays inherent in the internet TCP/IP protocols—
become constraining.® This data rate is currently around 4-8
Mbps (and hence most users will not notice an improved
browsing experience once data rates rise above this point).
Resolving this requires changes to internet protocols and
architectures—something that has to occur on an internationalbasis within internet standards bodies and key industrial
players.
Page Load Time (seconds)
1 2 3 4 5 6 7 8 9 10
Data rate (Mbps)
Source: Rosenthal 2014.
Figure 3.1: Page Load Times
There is also something of the-chicken-and-the-egg issue in that,
if data rates reach a point where the costs of increasing them
further are very high, then developers will be aware of this and
will not tend to develop applications that exceed this data rate.
Hence, the point where a major technical or economic step-
change is needed to increase data rates can become “sufficient”
in a self-fulfilling manner.
Alongside mobile coverage provided via cellular, there is
also mobile coverage provided via Wi-Fi. Indeed, estimates are
that of all the data provided to mobile devices, about 50 percent
to 80 percent’ is carried via Wi-Fi. Wi-Fi also typically provides
the connectivity in the home from the broadband access point.
Wi-Fi can, in the best-case scenario, deliver rates in excess of 50Mbps. However, these rates drop quickly if the Wi-Fi spectrum.
is congested or if devices are not close to the access point.
The sufficiency of the currently available speed was
confirmed in a recent study by the BCG [3] (Sherman et al. 2015)
which asked users about their mobile experience and
correlated it with the network speed and latency. They found
that once speeds exceeded 1.5Mbits/s there was no increase in
consumer satisfaction when watching video, and that
satisfaction when using apps stopped increasing at lower
speeds, below 1Mbits/s. They concluded that “the telcos’ race for
speed may, in fact, be a largely unnecessary endeavor that
breaks the cardinal rule of focused investment: spend where
the spending counts most.” They did notice that satisfaction
increased as latency fell, with latency levels as low as 75ms
appreciated for video and in the region of 25-50ms for some
other apps (4G currently delivers around 30ms latency in most
uncongested networks). They concluded that “cellular carriers
should, therefore, consider intensifying their efforts to reduce
network congestion.”
As BCG noted, current 4G networks deliver speeds well in
excess of 10 Mbps (when not congested). Indeed, some MNOs
now market speeds of above 100 Mbps, and manufacturers
have reported tests delivering over 1 Gbps when aggregating
multiple 4G carriers. It is very hard, then, to understand why
speeds faster than 4G might be needed. (The question of whether
more capacity is needed in order to ensure that speeds do not
slow as demand rises is considered next.)In passing, it is worth noting that there are parallels here
with the fixed broadband world. The same restrictions apply on
web browsing and broadly the same on video download. For
home usage, the highest data rate service is likely to be 4K
streaming video using around 20 Mbps. With multiple
occupants in the home, requirements might peak at around 60-
80 Mbps. This is well within the capabilities of fiber-to-the-
cabinet (FTTC) deployments which then use solutions such as
very high bit rate digital subscriber line (VDSL) or G.fast to the
building. The focus on delivering fiber to the home seems as
equally unnecessary as the high speeds within 5G. Further, it
might cannibalize investment that could otherwise occur in
important areas such as enhancing coverage, as examined in
later chapters. Indeed, in Australia where there has been
government-sponsored deployment of fiber to the home (FTTH),
the experience has been that (1) few subscribers opt for the
higher data rates available via fiber, with most selecting rates of
25 Mbps or 50 Mbps that would have been available on FTTC,
and (2) the additional time and manpower needed for FTTH
deployment has meant that homes that would have already
received an FTTC upgrade under an FTTC strategy are still
awaiting any form of upgrade. Here, the net result for the
consumer has been negative.®
Latency
As noted above, latency does have an impact on user
experience. Latency is broadly a measure of how long it takesfor a request to make it from the mobile phone to the end point
in the network. For applications like video streaming, latency is
irrelevant, as the handset can buffer material in advance and so
never needs to send an urgent request. But for applications like
web browsing, once speed is above 1 Mbps, it is latency that
determines the time it takes for a new page to load after the
user clicks on a link.
The importance of latency can be seen when web pages have
numerous elements. It is quite common for a web page to have
100 or so parts—text, images, headers, advertisements, and so
on. If the latency was 30 milliseconds (ms) and there were 100
elements all fetched sequentially, then the page load would take
3 seconds. In practice, some elements can be fetched in parallel;
web pages optimized for mobile use have much fewer elements.
Latency has fallen across the cellular generations. It was
around 500 ms with 2G, perhaps 100 ms with 3G, and around
30-50 ms with 4G. As BCG noted, falling from 100 ms in 3G to 50
ms in 4G has improved user satisfaction.
Further improvements may be both harder to deliver and
have less impact. As an example, most applications where
latency is seen as a critical issue involve video—such as VR or
remote control applications. Even the most advanced VR
headsets have a video refresh rate no greater than 100 hertz
(Hz), which means it takes 10 ms for a new video frame to
appear. The frame requires some processing before it can be
displayed. Hence, even with zero latency from the rest of the
system, around 15 ms latency is inevitable.The theoretical latency of 4G is 10 ms—additional delays
tend to occur within the core network of the MNO. Once the
message leaves the cellular network it may need to traverse
continents. The latency imposed by a message sent from the
East Coast of the United States that needs to reach a server on
the West Coast is around 30 ms, and the time for it to get from
Europe to the United States is around 60-100 ms. Some content
can be cached on servers within the same country, but this
typically only works for the most visited pages. So, a 4G
message might incur a 10 ms delay across the radio interface,
30 ms across the MNO’s core network, and 50 ms across the
internet. In this case, the radio delay is slightly over 10 percent
of the total. If the radio delay were reduced by half to 5 ms, the
total delay would fall from 90 ms to 85 ms, which would be
barely noticeable to any users.
Of course, core networks could be optimized to reduce the
30 ms delay, but this can be done with 4G networks perhaps
more easily than a complex virtualized 5G network, where
many of the resources in the core would not be under the
control of the MNO.
Currently, work is underway to reduce the theoretical 4G
latency to 5 ms by halving the size of resource blocks. Within
the 5G community, the original 1 ms latency target had been
determined impractically difficult and modified to 8 ms,
although these numbers may change as standards develop.
Latency is important, but 5G seems unlikely to have a
materially lower latency than 4G in practical situations. Even if5G did have a lower latency, the impact would be minimal,
since latency in other parts of the networks would dominate.
Predictions of Volume
The volume of mobile data has been growing ever since the
launch of the iPhone in April 2007. Figure 3.2 shows what
happened to both voice and data volumes in the years
immediately after the launch of the iPhone.
450 7
400
350
Voice
Petabytes/month
Data
21.02.03 G4 21 O2.03 24.21 G2. A3.04.Q1 22.0304 01
07 07 O7 O7 08 08 08 08 09 09 :09 09 10 10 10 10 11,
Figure 3.2: Growth in Mobile Data After the iPhone Was Launched
While voice grew steadily, data demand exploded, growing
around a hundredfold in the five years from 2007 to 2011. It has
continued to grow rapidly since then. Figure 3.3 shows the
predictions for mobile data traffic made in Cisco’s industry
renowned “virtual networking index” (VNI) in recent years.?