About Bloobble

Random Presentation

.: Try it :.

NEW! CATEGORIES

I am just starting this, but I wanted to share before getting all categories done. Thanks to send me your feedback and comments.

Study to Assess Broadband Bandwidth Usage and Key Trends in Europe

Enjoy this presentation?

You are not logged in. You will be automatically directed to the login screen if you want to download, bookmark or share this presentation.

These are available as part of the subscription package
May be the time to ckeck out a subscription!

These are the first of many nice features we plan
on adding to save your time ::

Get links

Just push the Select button corresponding to your choice, CTRL + C to copy and then use Paste in your target application.

Link to this presentation:

Embed in your web site:

Presentation Transcript:

insight from data / knowledge into action
Study to Assess Broadband Bandwidth Usage
and Key Trends in Europe
February 2008
Produced independently by Ventura Team LLP for the
Fibre to the Home Council Europe.
Executive Summary
In this study we have examined broadband traffic patterns over several years at the European level and for a sample of
100,000 true broadband homes (i.e. those providing consumer services of 100Mbit/s or faster) in Europe.
Our first objective was to test Nielsen’s Law for Europe. In 1998 a US based web page usability consultant called Jakob
Nielsen postulated a `Law’ of bandwidth that projected a 50% increase every year in the broadband speed available to a
high end customer. This was similar in concept and formulation to Moore’s Law for the computer industry but slower
(Moore’s Law implies 60% pa increase in computing power).
We find that Nielsen’s Law on average has held remarkably well for more than 20 years although Government
intervention in Sweden lead that country to greatly exceeded the rate of progress predicted by the Law. Using this
“Law” as a guide we expect 100 Mbit/s services for high end consumers in the France in 2008, Poland in 2012 and
the UK in 2015. Such services have been increasingly common in Sweden since 2004 but Europe is, sadly, way
Page 2
the UK in 2015. Such services have been increasingly common in Sweden since 2004 but Europe is, sadly, way
behind Japan and Hong Kong which have had 1 Gbit/s services for a couple of yeas already.
Our second objective was to see if existing true broadband customers generate more traffic than conventional
broadband users. If they do then it shows that the high speed of true broadband is indeed useful to the customer.
The data shows that true broadband homes do generate significantly more traffic than similar homes on ADSL.
Typically this is 3x the traffic per home. As higher speed access and high definition video becomes more common
we expect traffic per home will grow and probably faster from fibred homes.
We expect this to grow driven by larger screens, HDTV even higher resolution video and in time new applications.
The message is that consumers want (and despite today’s rudimentary technology already use) the greater
capacity of true broadband. In future we believe they will demand even more.
True broadband networks enable entirely new patterns of video consumption as already evidenced by the Swedish
experience (we believe that it is no coincidence that major filesharing resources and privacy proponents are in Sweden
where true broadband first reached critical mass in Europe). The growth of FTTH seems likely to generate considerable
change in the audio-visual sector over the next seven years.
Contents
• Objectives and MethodologyPage 4
• Testing Nielsen’s LawPage 8
Page 3
• Is Demand per Home Already Growing?Page 18
• Comparison of FTTH v ADSL Broadband
Traffic Patterns
Page 22
• Conclusions and Future Growth DriversPage 30
Page 4
OBJECTIVES AND METHODOLOGY
Measure actual traffic from true broadband homes. Compare to ADSL.
We Had Two Objectives
The Fibre to the Home Council Europe is a trade body concerned with promotion of investment in fibre to the home
(FTTH) networks. Building such a network is very capital intensive and the question of why consumers would want FTTH is
a pressing issue both for the telecom industry and also increasingly for Governments that want to ensure their own
population and businesses have an up to date modern telecom infrastructure to enable success in the new digital age.
Our first objective was to test Nielsen’s “Law” which was postulated in 1998. Nielsen’s Law of bandwidth is similar to that
of Moore’s Law for the number of transistors on a silicon chip. It states that the bandwidth available to a high end user
grows 50% per annum on average and that the mass market lags the high end user by 2-3 years.
Our second objective was to examine current world observed usage behaviour by those fortunate enough to have true
Page 5
Our second objective was to examine current world observed usage behaviour by those fortunate enough to have true
broadband (100 Mbit/s or more). Do looking at what such customers are doing today we wanted to see what evidence
there already for the continuation of Nielsen’s Law (or something like it) into the fibred future? We wanted to so this
because when discussing FTTH business cases and investments, two basic questions about bandwidth usually arise:
Who needs all that bandwidth?
What will they use it for?
Note: Ventura Team LLP consults regularly in FTTH business strategy, project financing and interim management. We are
a consulting and angel investment Partnership of several senior industry executives (from fixed and mobile backgrounds)
and investments in various operator and service provider companies including in FTTH. We are best contacted via
info@venturateam.com
We Chose Sweden as the Focus
The Swedish FTTx market is the most advanced in Europe and the third most fibred nation in the OECD after Korea and
Japan. Some years ago the Swedish Government pump primed roughly $4 billion of open fibre infrastructure investment
which was amplified by Municipalities and municipally owned utilities.
Around 2/3 of primary residences in Sweden are apartments which are more economic for fibre deployment than stand-
alone houses.
30
35
Other Fibre/LAN Cable DSL
OECD Broadband subscribers per 100 inhabitants, by technology, June 2007
Page 6
0
5
10
15
20
25
30
Source: OECD
OECD average
Methodologies
Nielsen’s Law
To examine the applicability of Nielsen’s Law we gathered historical information about the launch of new
broadband services for high end consumers. To do this we studied a variety of sources (press reports, old investor
presentations) and interviewed a number of senior industry contacts.
In each case we determined through our research the first the year of the first reasonably wide commercial
availability in at least the major city (ideally several cities) of a new broadband speed consumers service.
We then plotted these results in log space using the same base data for dial-up speeds that Nielsen himself used (by
its nature the telephone network and the modems of that era had very similar performance regardless of country)
Page 7
Traffic on fibre networks
This was the most difficult part of the study. We managed to obtain traffic information at various levels of detail
from several operators and discussed patterns of usage with a larger group by means of telephone survey. We also
examined what records exist in the public domain for traffic flows and patterns in the countries concerned and used
interviews or published sources to determine the size and type of broadband customer bases for the different
operators.
By focussing on Sweden we were able to assemble a reasonably consistent dataset at the level of detail needed for
this study from a base of 100,000 true broadband homes spread across various operating companies and cities. This
is a large enough sample to be statistically significant when compared to ADSL in the same national market. To
varying degrees through the interviews and desk research we were able to validate the general conclusions as being
valid for the other countries.
Page 8
TESTING NIELSEN’S LAW
Does This 1998 `Law’ Still Work?
Nielsen’s Law for Bandwidth was Similar to Moore’s Law for Silicon
In 1998 the web was fast growing but very new still and
Jakob Nielsen was concerned with usability of web
pages. Web designers tended to get carried away with
the new medium and made pages too cluttered for the
slow connections of the time. Nielsen’s point was that
over dial-up it would take several seconds to load each
page which was frustrating.
Looking at his own Internet access speed (see top right)
he postulated a Law and suggested that it would not be
until 2003 that people could enjoy complex pages.
Source:theoriginalcurvefromNielsen’spage
www.useit.com/alertbox/980405.html
Page 9
until 2003 that people could enjoy complex pages.
His Law had two parts:
A high-end user's connection speed grows by 50%
per year (this is 7.5x after 5 years)
The mass market lags the high-end by 2-3 years
His Law implied a slower rate of progress in bandwidth
than the capability of computing – a trend which has in
fact been realised as reported by Level 3 for the USA
(see bottom right).
CAGR Compound
over 10 years
Nielsen Bandwidth 50% 57x
Moore Computing 60% 100x
Source:Level3InvestorConference2007
basedonvarioussources
Nielsen’s Law Works Perfectly for the UK
Our analysis of the UK’s broadband history provides the
speed curve shown on the right.
Note that the vertical scale is logarithmic so it is non-
linear. The higher up the vertical scale the greater the
difference – each line represents a 10 fold increase in
the speed available to a high end user.
The trend rate of progress is almost exactly the 50% per
annum predicted by Nielsen (in fact it is 49.87%).
The curve predicts that 100 Mbit/s will be available in
2015 and 1 Gbit/s in 2020. Given the planned
Page 10
2015 and 1 Gbit/s in 2020. Given the planned
investment in VDSL and cable TV upgrade and the
availability of LANs in many apartment blocks this
seems credible bearing in mind that Nielsen refers to
“high end Internet users” who would tend to live
clustered in certain high income urban neighbourhoods
with good telecom infrastructure.
Note: Some University residences and even multi-tenant
office blocks are already operating at 100 Mbit/s but
those are small private systems and not relevant for this
analysis.
The Rate of Progress was, Unsurprisingly, Variable
This chart shows the annual rate of change in our data
set for the UK. We do not have results for every year as
the industry tends to progress by step change (so in
some years the highest speed available was the same as
hat for the year before).
This is shown on the chart right where:
The wavy line shows the compound annual growth
rate between each observation in our data set. In
other words if speeds trebled after 2 years then
the CAGR would be 3^(1/2 years)=73% each year
Page 11
the CAGR would be 3^(1/2 years)=73% each year
The straight line shows the trend of the data
(which for the UK conforms to Nielsen’s predicted
50%) which is, of course, the same each year so
the line on the chart is flat.
Sweden Beat Nielsen’s Law
We used the same dial-up era data but plotted the
curve for Sweden’s early deployment of Ethernet
networks.
The first 10 Mbit/s service became available in limited
areas in 1998 and this is the date we have used. There
is no hard and fast definition of geography in Nielsen’s
Law but by 1998 we know that handful of ISPs and cable
operators were offering 10Mbit/s service and that by
1999 the trend was becoming established in many
areas.
Page 12
areas.
Similarly by 2004 100 Mbit/s was available in several
areas of the country.
The curve implies a high-end 1 Gbit/s service will be
marketed in Sweden this year or next year. Given the
simplicity of the Gigabit interface, the existing extensive
residential fibre and first deployment of PONs and -
most importantly - growing competition in the fibre ISP
space, then this seems entirely plausible.
Sweden is 10 Years Ahead of the UK
As a way of deriving a simple message from the
complicated empirical data, we plotted the trend lines
for Sweden and the UK on the same chart.
Looking at the positions for of the two countries for 100
Mbit/s deployment and for 1 Gbit/s we can see that
Sweden leads the UK by about 10 years (or perhaps
more).
This is not a “scientific” conclusion and depends on the
UK performing as poorly as predicted by what is after all
a rough guide. No doubt protagonists will raise any
Page 13
a rough guide. No doubt protagonists will raise any
number of detailed objections but it is clearly the case
that Sweden is ahead and Nielsen’s Law gives us a
useful, albeit simplistic, way of estimating by how much.
France is Just Ahead of Nielsen’s Law and Beats the UK
Page 14
France has made steady progress and beats the Law by a modest margin.
Following a flurry of FTTH investment announcements last year we have assumed 2008 is the year that enough 100
Mbit/s services become available to be meaningful. Even if delayed to 2009, France has been doing better than the UK
and seems set to continue to do so.
Spain Almost Matches Nielsen’s Law
Page 15
Our research indicates that Spain is very close to Nielsen’s Law. 46% CAGR we estimate compared to the predicted 50% is
very close and the margin of error is probably greater than this difference.
We may have missed some recent announcements. We are also aware of smaller independent networks and it may be
underestimating their services.
Poland is Doing Well but Illustrates a Simplification in Nielsen’s Law
Page 16
Poland has extensive cable TV networks and also ~700,000 homes on local Ethernet networks though most of these are
not yet connected to the Internet with fibre. There has been a great deal of activity in Warsaw and major cities to
increase speeds (particularly by cable TV operators) and for new upmarket apartment developments to be provided with
high speed services as part of the construction process.
On this basis Poland beats Nielsen’s Law and does almost as well as France. However the disparity between services in the
cities and the Polish countryside is huge and we believe much greater than in France, the UK etc. The reader should
recall that Nielsen’s Law applies to high end users who, almost by definition, are in the locales with the best
infrastructure. It says nothing about the equality of service throughout a country.
An Aside – An Organisation Specific Example from the USA
The University of Cedarville publishes a long run
history of its connectivity on its website
On discovering this we decided to test Nielsen’s Law at
the level of a an organisation. Clearly the Law would
be unreliable here because as well as variability in the
telecom industry’s ability to supply locally (more
variable than nationally) academic budgets and many
other factors come into play which fall away from a
national level analysis.
Despite this the results are of the right magnitude to
Page 17
Despite this the results are of the right magnitude to
fit Nielsen’s Law which, after all, was intended as a
rough guide rather than a precise prediction. The
speed of connectivity at Cedarville grew 69% (trend
line 61%) between autumn 1993 and autumn 2007
(annual data).
Page 18
IS DEMAND PER HOME ALREADY GROWING?
Have improvements in ordinary broadband speed had an impact on usage?
Has Bandwidth Usage by Individuals Grown?
Although Nielsen’s Law indicates that broadband speeds have risen
and will rise further, this does not necessarily mean that actual usage
in terms of the traffic volume from each home has grown at the same
rate (or indeed grown at all).
We have not been able to obtain detailed long run time series for a
suitable set of ISPs to answer this question but we have found two
international anecdotal examples (see right) and also have examined
overall international bandwidth trends. International bandwidth is a
separate market within the telecom industry and its growth has been
documented over many years by Telegeography.
Page 19
documented over many years by Telegeography.
Telegeography’s various publications show roughly a 50% per annum
rise in international bandwidth for Europe (see top right) and in 2005
they also stated that “While the recent growth rates of Internet traffic
are substantial, they do not approach the `doubling every 100 days’
calculation that was so often cited in the 1990s. TeleGeography
estimates that international Internet traffic will double approximately
every two years.”
Various studies suggest that usage has grown as the two examples
from New Zealand and the USA shown right suggest.
The question is that although total traffic has risen so has the
number of broadband homes in Europe so has usage per home
actually risen after broadband first arrived?
TheExabyteEra,reportbyCiscoSystems,Jan2008
We Estimate Traffic per Broadband Home Grew at ~20% CAGR 2002-07
1) We obtained the numbers of both dial-
up and broadband connections in Europe
(red and blue columns respectively)
2) We then estimated traffic from a variety
of sources and developed a total volume
index (the green line) showing the relative
amount of IP traffic from and between
European countries relative to the base
level in 2002. This was cross-checked
against other sources (see next slide).
Page 20
against other sources (see next slide).
3) By adjusting for the mix of dial-up and
broadband we estimated an indicative
change in usage per broadband connection.
We believe average usage per broadband home grew almost 20% per annum over this period
(about 250% in total) generated by a combination of higher line speeds and more visual web
pages and other bandwidth intensive types of use.
Our analysis assumed that 1) the level of dial-up traffic per user did not materially change after 2002 on average 2) business usage stayed constant
relative to consumer usage and 2) international traffic reasonably reflects growth in both total domestic and international traffic.
We believe that these are reasonable assumptions in this context.
Note: Our Underlying Estimates Seem Consistent
with Observed Traffic Growth in Europe
The best sample of whole Europe traffic growth
comes from www.telegeography.com
Their figures are shown left.
If we assume that this represents general growth
reasonably well then we can cross-check and guide
our bottom-up estimate to match the pattern of
growth (bottom left). The two match very well.
As a further cross check we looked at the main UK
internet peering exchange LINX as representative
of a middling speed ISP market. The total traffic
SourceTelegeographyaspresentedatPTC
Page 21
of a middling speed ISP market. The total traffic
growth over the period for LINX was 8x (green
shaded area) – remarkably close to our estimated
figure for Europe as a whole of 8.2x.
Source
Source:www.linx.net
capturedFebruary2008
Page 22
COMPARISON OF FTTH V ADSL
BROADBAND TRAFFIC PATTERNS
More – much more!
We Compared True Broadband to ADSL in Four European Countries
Our main research focus was Sweden because there we could examine data from around 100,000 fibred homes – a far
larger and representative sample than is possible at the moment anywhere else in Europe.
We cross-checked our findings with fibre speed operators in three other countries. There were some differences
compared to Sweden but all true broadband operators seem to experience significantly higher traffic per home compared
to their local ADSL counterparts.
The detail and specificity of the traffic data that we were able to obtain varied considerably by operator but the main
comparative result was made at the aggregate level and we believe it is reliable and as accurate as any such estimate can
be. Furthermore, the findings were confirmed and supported by the experts and managers that we interviewed across a
wider set of operators in other countries.
Page 23
Fibred Homes Generate 3x the Traffic per Home of ADSL
Inbound & Outbound Traffic – 24 Hour Mid-Week Average
The columns show the average throughput (as Mbit/s per
customer) measured for 100k homes across 4 pure fibre
operators labelled A-D. Although they would have some
small businesses, most of their customers are homes.
The 5th column shows the average for ADSL customers in
Sweden. The red line shows the fibre average.
Inbound and Outbound traffic for Pure Fibre OpCo’s
is 3.2 to 4.4 times higher respectively compared to
Inbound Traffic Mbits/sec/Sub - 24 Hr Average
Fibre Average
0.000
0.050
0.100
0.150
0.200
0.250
Mbits/sec/Sub
-
0.1
0.1
0.2
0.2
0.3
3.2 Times-X ADSL Av
Page 24
is 3.2 to 4.4 times higher respectively compared to
ADSL average suggesting both higher streaming and
P2P file sharing.
Anecdotal experience from some Fibre OpCo’s is that
they are net exporters of packets driven by
filesharing, but our sample suggests a similar
Inbound / Outbound Ratio for both Fibre & ADSL.
Over the 24-hour period true broadband usage is over 3-
time more which is significant at this stage of market
evolution given that many of the mass market applications
have not been “boosted” to realise the advantage of Fibre.
ADSL average was devised from operators with predominantly ADSL, with limited / No Fibre.
0.000
A B C D ADSL Av.
-
Outbound Traffic Mbits/sec/Sub (Av) - 2200 - 00 Hrs
Fibre Average
0.000
0.050
0.100
0.150
0.200
0.250
A B C D ADSL Av.
Mbits/sec/Sub
-
0.1
0.1
0.2
0.2
0.3
3.4 Times-X ADSL Av
The Pattern is Slightly Higher in the Peak Hour
Inbound & Outbound Traffic – 22.00-00 hrs Average (Mid-Week)
Late evening is peak time for broadband use in most
European countries. We looked at the ratios for the 10pm
to midnight period to see if there was any difference in
peak time behaviours.
Inbound and Outbound traffic for Pure Fibre OpCo’s
are both approximately 3.8 times greater than ADSL.
Outbound traffic for a Pure Fibre OpCo during the
Peak-Period is 3.85 time greater than the ADSL
Inbound Traffic Mbits/sec/Sub (Av) - 2200 - 00 Hrs
Fibre Average
0.000
0.050
0.100
0.150
0.200
0.250
Mbits/sec/Sub
-
0.1
0.1
0.2
0.2
0.33.84 Times-X ADSL Av
Page 25
Peak-Period is 3.85 time greater than the ADSL
Average, which is slightly greater than the 24-hour
average (3.4 times).
Inbound/Outbound Traffic Ratio for Pure Fibre OpCo’s is
essentially the same as for ADSL suggesting that end-users
are using the same applications but on a larger scale.
However even in Sweden it is early days for consumer true
broadband - we would expect this to increase significantly
as fibre become more main-stream as applications are
developed with Fibre in mind.
ADSL average was devised from operators with predominantly ADSL, with limited / No Fibre.
0.000
A B C D ADSL Av.
-
Outbound Traffic Mbits/sec/Sub (Av) - 2200 - 00 Hrs
Fibre Average
0.000
0.050
0.100
0.150
0.200
0.250
A B C D ADSL Av.
Mbits/sec/Sub
-
0.1
0.1
0.2
0.2
0.3
3.85 Times-X ADSL Av
The Differences Are Less Marked but Still Significant in the Early Morning Trough
Inbound & Outbound Traffic – 0400-0600 hrs Average (Mid-Week)
4am to 6am is the trough in usage in Sweden. Most people
are in bed and many backups or downloads have finished
so traffic at this time if night represents the background
load of large fileshares and automatically driven traffic
perhaps to the USA which is in its evening peak.
Inbound and Outbound traffic for Pure Fibre OpCo’s
is 2.98 to 3.04 times higher respectively compared to
ADSL average suggesting both higher streaming and
Inbound Traffic Mbits/sec/Sub (Av) - 0400 - 0600 Hrs
Fibre Average
0.000
0.050
0.100
0.150
0.200
0.250
Mbits/sec/Sub
-
0.1
0.1
0.2
0.2
0.3
2.98 Times-X ADSL Av
Page 26
P2P.
Inbound/Outbound Traffic Ratio for Pure Fibre Operators
is again essentially the same as for ADSL.
Regardless of time of day, it is not clear whether P2P
applications will significantly shift Inbound/Outbound
Ratio as this depends upon the distribution of active peers
and leeches in the filesharing network.
Bit –Torrent style protocols are evolving to better support
streaming which will potentially offset any increases in
Outbound Traffic due to P2P applications.
ADSL average was devised from operators with predominantly ADSL, with limited / No Fibre.
0.000
A B C D ADSL Av.
-
Outbound Traffic Mbits/sec/Sub (Av) - 0400 - 0600 Hrs
Fibre Average
0.000
0.050
0.100
0.150
0.200
0.250
A B C D ADSL Av.
Mbits/sec/Sub
-
0.1
0.1
0.2
0.2
0.3
3.04 Times-X ADSL Av
Why Might Fibred Homes Generate More Traffic?
Traditional ADSL Traffic Pattern
The traffic measurement below is taken from mid-sized ISP in a
market which has a reasonable and still growing penetration of
broadband. Figures show traffic in both directions for a typical day in
December 2007. Browsing (inc flash from YouTube etc and http
download) dominate the traffic.
Example of Consumer Fibre Network, Morning Weekday
Off peak, a lot of traffic seems to be machine driven by peer to peer
filesharing as home media servers sit there gathering the evening’s
entertainment. Also filesharing clients worldwide on ADSL
automatically seek out the higher speed uploads available from those
with fibre (so some fibre networks actually export traffic overall which
is unusual for a consumer broadband network).
Page 27
Note: In both cases entertainment is THE driver of demand.
The differences lie in the detail of service mix and content distribution.
While Clearly Important, We Believe Filesharing is Not the Only Factor
TCP/UDP Prot
ocol
Data Flows
Accumulated Percentage / Historical Protocol
View
dns 7.0 GB 55,048 4.4%
ssh 4.1 MB 573 0%
http 18.1 GB 371,472 11.2%
ftp 570.6 MB 5,698 0%
mail 48.5 MB 15,080 0%
voip 5.0 MB 9,675 0%
directconnect 16.8 MB 968 0%
bittorrent 5.7 GB 49,542 3.5%
kazaa 88.8 KB 56 0%
edonkey 12.2 MB 850 0%
gnutella 99.2 KB 113 0%
msnmessenger 1.8 MB 597 0%
Other TCP/UD
P-
based Protocol
129.6 GB 4,957,699 80.5%
This chart illustrates the problem of measurement.
As p2p traffic is designed to work around firewalls
and is increasingly encrypted etc it can be difficult or
impossible to identify without expensive equipment.
This data is from Sweden and the interviewee told
us that the “Other” category shown left will be
almost exclusively p2p traffic of various types.
However this may however be a peculiarly Swedish
result. In other countries respondents believed that
Page 28
based Protocol
s
129.6 GB 4,957,699 80.5%
Accumulated
View
Historical
View
Note:
• What is a flow?
o TCP: a flows is a TCP connection.
o UDP: a flow is a packet.
• TCP flows are not accounted for fully (sender and recipient) remote peers.
result. In other countries respondents believed that
video streaming was an increasingly important
driver of traffic and that filesharing was in some
cases “no more than 50%”.
They did, however, confirm 2x-5x greater traffic
levels compared to ADSL.
We briefly discuss future trends in and drivers of
traffic in the next section.
Example of Benefit - Low Latency for Gamers
These Differences May Seem Slight but Affect the Purchase Decisions of Opinion Formers and so
Do Impact the Market
200
250
300
200
250
300
Sources of Delay in Man-machine Gaming Response in Milliseconds
ADSL is 50% slower FTTH (with fibre grade CPE)
Page 29
0
50
100
150
fastest right
handed eye-mouse
movement
clocked processor
and fast screen
refresh
ADSL firewalls ADSL line delay
0
50
100
150
fastest right
handed eye-mouse
movement
clocked processor
and fast screen
refresh
ADSL firewalls ADSL line delay
Page 30
CONCLUSIONS AND FUTURE GROWTH DRIVERS
“There are no speed limits on the road to excellence.”
Can Nielsen’s Law Continue to Indicate the Future?
Nielsen’s Law was only ever intended to persuade web
designers to restrict the complexity of their web pages
until a reasonable proportion of US homes had
broadband which he expected to happen around 2003.
Despite this humble origin, a decade later it is still
working fairly well as a guide to the trend in broadband
speeds even in Europe.
Can this ~50% pa rate of demand growth continue? We
believe that it will for at least another decade driven by
various factors including:
Video
walls &
digital
home
Page 31
various factors including:
Increasing file size to support consumer
electronics improvements
People spending more time on the Internet
Diffusion of large screens going beyond HDTV and
the gradual evolution of the digital home
Immersive gaming and perhaps new applications –
who predicted the huge popularity of social
networking before myspace.com?
File size
More
time
online
Filesharing Has Pushed Up Average Traffic per Individual
As More Video Comes Over the Web and Files Get Larger this Trend Should Continue
AspublishedbyTelegeographyin
conferencepresentation
http://net.doit.wisc.edu/data/flow/size/
Page 32
The chart above left is of historical interest because it
shows the impact on flow sizes (and implicitly traffic
volume per user) over a one year period at the University
of Wisconsin in the 12 month period when Napster
suddenly became popular. Mp3 files were typically larger
than the file types consumed in a web browsing session at
that time.
Napster has long since faded away as a major drive of
Internet traffic but today the music charts are largely
determined by Internet downloads rather than sales of
CDs. As broadband speeds rise the pattern of video seems
set to undergo an analogous transformation, generating
much more traffic as a result.
The chart above right from Cachelogic shows the rise of p2p
filesharing traffic over a decade to dominance of volumes
on mass market broadband ISP networks. Our survey of
Swedish fibre ISPs indicates the p2p is nowadays an even
higher proportion of traffic in an advanced market.
The number of customer using filesharing remains unclear
but anecdotal evidence suggests it is still a technology for
the young and savvy that are prepared to pirate content. If
filesharing becomes more mainstream, perhaps as the
amount of legal content increases and ease of use of the
software improves then it clearly has the potential to
generate huge traffic volume.
The Internet Has Already Partly Substituted for Traditional TV
The time budget and viewing hours
result shown right is for the USA but
the pattern will be similar in Europe
we believe. Consumer interest and
the advertising spending that it drives
are shifting from TV to the Internet.
In response many European
broadcasters are moving online in
various ways and there is a quest for
viable video business models. At the
Page 33
viable video business models. At the
same time the media industry is
taking the first steps to co-existing
with the reality of file sharing (cf
recent content deals with BitTorrent)
The stage is set therefore for video
consumption over the
Internet, however it works in
detail, to grow steadily and drive
traffic volumes ever higher.
TV Screen Sizes are Increasing 20% Per Annum – Blu Ray and Ever Higher
Definition Video Will Mean Larger Files
This measurement of browser screen resolutions
shows a slow but steady increase in screen resolution
as PC hardware improves of around 10% each year.
A similar compound rate of growth applies to
consumer TV image resolutions (as measured by
pixels) over the last 30 years (below right)
However, the increase in television screens is much
faster at the moment as a result of the replacement of
CRT displays by LCD and plasma. These new
technologies are completely reshaping (literally)
http://www.w3schools.com/browsers/browsers_display.asp
Page 34
technologies are completely reshaping (literally)
television.
According to display market intelligence experts
Meko, the weighted average size for a TV set in
Europe will increase from around 26" at the beginning
of 2006 to just over 34" at the end of 2009. This is a
CAGR of 20% (by screen area) and with increasing
sales of 70 inc screens and the launch of 80+ inch next
year this trend seems set to continue over the
medium term.
Broadband will need to support larger and larger file
sizes to fit high definition programming and
larger, higher resolution screens
http://www.w3schools.com/browsers/browsers_display.aspSource:VenturaTeamderivedfromWikipediadata
Cisco’s Detailed Bottom-Up Projections Assume Entertainment Drives
Volume Growth per Home (augmented by new apps and mobile)
Cisco’s Global IP Traffic
Forecasts are based on very
detailed bottom-up analysis of
drivers and trends including
assessment of historical
behaviour.
Their model projects growth in
European traffic volume of 8x-
9x over the 5 year period 2006-
2011
Page 35
2011
This is roughly consistent with
our own analysis of past trends
and equates to a CAGR of
~50% per annum
Source: Global IP Traffic Forecast and Methodology 2006-2011, Cisco, 2008
Key Messages
This study has shown that:
As a general guide, European broadband speeds are rising around 50% per annum and should continue to do so.
This is Neilsen’s Law in action.
When customers have faster connections they use them more. This is true for conventional broadband and even
more so for fibre – in fact fibre customer use broadband (in terms of traffic) about 3 times more than ADSL
customers and it is quite possible that this difference may grow further over time because it is still very early days
for the fibre speed Internet.
It takes time for the higher speeds to spread from the aficionado to the mass market but they do and traffic keeps
growing. Broadband traffic per home has been growing at 20% pa over the last few years and this will continue and
the growth rate will probably accelerate significantly both as video over the Internet becomes more common and
Page 36
the growth rate will probably accelerate significantly both as video over the Internet becomes more common and
also more homes are provided with true broadband.
We believe that these trends have a number of implications:
Operators will need to invest and upgrade to take true broadband into the home. Eventually customers will simply
require this but the most rapid progress will be made in markets with high infrastructure competitive pressure or
where a regulator acts so as to simulate such pressures.
Investors will need to understand the nature and potential operational / regulatory and financial structures of this
€100 trillion wave of capital investment in fibre access across Europe.
Governments (local and national) have two issues to consider – in countries which are only on or below Nielsen’s
curve, are they happy to trail behind competing nations and of not, what is the best pump priming or other action
they can take? Within country, how great a divide between the best served areas and the worst is tolerable and
what can be done to minimise it.
Customers can look forward to more speed and bigger better pictures while media companies face a period of
continuing change and need to respond to the impact that high speed Internet distribution and ever falling local
storage costs are having on consumer behaviour and content distribution..