bakar optika pristupna mreza
TRANSCRIPT
-
8/13/2019 Bakar Optika Pristupna Mreza
1/101
2.1 Technological BackgroundFigure 2.1 gives a high level overview of the network. One can readily distinguish different structures in the
network. The separate structures have important differences in traffic, equipment, technologies and operations.
Different astractions and functional or logical separations help in reducing comple!ity in each part. Theoperator maintains a good overview of the whole network and technicians can more easily work at one separatepart without much concern of the other parts. Different such astractions for the network e!ist.
Figure 1High level overview of the network
Thephysical structure contains first the telecom specific infrastructure, all infrastructures required to provide the network
connectivity and functionality to the customer. This contains all uildings for hosting network equipment, distriuted over thefull surface of the network. "t contains all street cainets, poles, wires, uried cales and conduits, wireless ase stations, etc.#t the network side it might as well contain the customer equipment, depending on whether this is considered under the
responsiility or ownership of the operator. The physical infrastructure will also contain other non-telecom specificinfrastructure such as all operational and research uildings, all equipment at this side, etc.
$onsidering the operational structure, the telecom specific operations concern the network and the customers. Thisincludes maintenance, repair, planning, pricing and illing, customer resource management, etc. Non-telecom specific
operations concern all operations for management, sales outside the telecom usiness, the financial department, etc."n oth the physical and operational structure, not all parts should e owned and operated%performed restrictively
y one operator. # network operator can outsource parts of the operations over the network to the equipmentvendor or to another party. #lso the infrastructure might e partly leased to%from another network infrastructure
operator. Often the regulator will demand parts of the network of the incument operator to e opened up, for a
fair price, to other licensed operators &O'O(. # little more on this can e found in chapter ). This dissertationwill focus on tier * and in e!tension tier 2 operators. The focus is further restricted to the lowest functional
layers, especially layer 1 and in e!tension up to layer *. Finally the dissertation will also look into more detail to
the access network and only secondary look at metro and core networks &as e!tensions descried in chapter )(.The main study of this dissertation will consider all physical and operational structures required for providing
connectivity to the customer up to the point where layer 2 or layer * functionality and equipment is provided. "twill neglect all physical and operational structures which are non+telecom specific.
1. 2.1.4 Access NetworkThe access network transports and aggregates all traffic of the customers up to the metro network. nlike in the
metro and core network, the access network will typically only direct the traffic to the first metro node and notuse any or very limited routing in the network. The access network has a tree structure via which it directs all
traffic to the root at which the traffic is handed over to the metro network. There are various technologies in use
in the access network, which are typically split etween wireless and wired line technologies. This dissertation
will focus on fi!ed access technologies and more in detail on the deployment of the novel FTT- technology.Figure 2.* gives an idea of the ma!imum andwidths offered y various wired line technologies and it is clear
that FTT- offers andwidths not easily attainale with other e!isting and even future access technologies. Thefigure also shows the increase in access andwidth over time taken from the and the /wedish market and fits
an e!ponential to oth increases &for the this corresponds to 0ielsens law( 2.34. 0ielsens law predicts ayearly increase of 567 or a tenfold increase every si! years. "f we assume this law to hold for the ne!t decades,
then the andwidth to the customer would reach a andwidth over 1Tps y 2686.
-
8/13/2019 Bakar Optika Pristupna Mreza
2/101
Figure 2 Bandwidth evolution in the access network
9hat follows is first a very short introduction to wireless access technologies and future
development. #s the core of this dissertation focuses on fi!ed access networks, we will
descrie in more detail the fi!ed access technologies in use today. #s a final susection of this
fi!ed access description, FTT- as the ne!t step in the evolution of fi!ed access is introduced.
# more detailed description of its evolution is postponed to chapter *.
2.1.8. Fi!ed #ccess 0etworks
9ithin the fi!ed line &or wired( access networks, digital suscrier line &D/'( and hyrid fire coa! &-F$( arethe most prominent technologies. Fire to the home allows much higher andwidths than other fi!ed access
networks and is currently eing rolled out in various countries as the ne!t generation access networkinfrastructure.
Digital /uscrier 'ine
Originally copper ased networks were deployed in order to provide telephony service to a given region. #s such
they are widespread, for instance in :elgium they cover almost 1667 of the households. 0e!t to voice, thesecopper networks also allow transporting data, and first modems provided data transport in the voice and up to
55kps &1;)6+1;;6(. Digital suscrier line &D/'( uses the higher frequency ands to transport data, and reaches
much higher andwidths up to 166
-
8/13/2019 Bakar Optika Pristupna Mreza
3/101
Figure 3xDSL technologies - data rate vs. reach
Figure 4 Overview of a digital subscriber line network
The D/' modem at the right of the figure connects to the twisted copper in+house wiring. The in+house wiring connects to the
outside plant y means of a wall plug. The incoming data is separated from the voice efore the telephone y means of a splitter.:oth voice and data are transported over the outside plant to the first active point. The outside copper plant has a continuous
dedicated twisted copper pair connected at aggregation points and fle!iility points such as the street cainet &/$(. The first active
point is called the D/' access multiple!er &D/'#
-
8/13/2019 Bakar Optika Pristupna Mreza
4/101
Figure 5 Overview of an FTTH network
The optical network termination &O0T( at the customer side performs the translation of the optical signal to the
in+house wiring. 0e!t to =thernet it is not uncommon to perform a translation to a road range of e!istingconnectors at this point &coa!, twisted pair, wireless(. The inside optical wiring is connected to the outside plant
y means of an optical connector plug often referred to as the optical network termination point &O0T?(. Theoptical signal is transported over the outside plant up to the central office. "n the outside plant there are different
aggregation points ?( and often there is also a fle!iility point &F?( comparale to a street cainet in copperased networks, closest to the customer. #t this point there are various options for the telecom operator. ?assive
optical networks &?O0s( will aggregate the optical signal of different fires into one fire at such aggregationpoints using passive optical splitters and as such create a point to multipoint network with the optical fire as
shared medium. #ctive optical networks O0s( will connect the customer with a dedicated fire up to the O'T.#lso the numer of customers per ?O0, #? and F? are degrees of freedom for the operator installing the
network. #t the central office, all optical fires connect to the ODF and from there to the optical line terminalwhich will aggregate all traffic, and translate etween protocols where necessary 8. #t this point one or more
additional wavelength&s( can e used for roadcasting content &for instance F video( to all customers of a ?O0.:eyond the O'T, the traffic is sent into the metro network. #n e!tended description of the different
architectures, topologies and technologies can e found in chapter *.
&8
"n case of a ?O0, access to the shared medium is divided etween the different customers y means of some division
multiple!ing ased protocol. $urrently most often time division multiple!ing is used in which each customer gets a smallrepetitive partition of the time during which he transmits and%or receives his data(.
There is currently no large scale FTT-+network deployed in the :elgian market. :elgacom announced they wille performing FTT- trials at the end of this year 2.264. The FTT- scene in =urope is also very small, withsome countries in which FTT- is rolled out fast &especially /weden(. On a world scale, FTT- is prevalent in
#sia, especially in Eapan and orea. "t is deployed on a much smaller scale in / however the / still has amuch larger customer ase than =urope &2.5
-
8/13/2019 Bakar Optika Pristupna Mreza
5/101
the results &and process(. /uch acting is rather ovious and not in direct relation to the construction of a usiness
case and is as such left out of the methodology.
Figure 6Methodolog used for the Techno-!cono"ic research
2.).1 2.2.1 Design ?hase"n a first step, we delimit the scope of the usiness case. # plan is made of what will e the aim of the usinesscase, what will e taken into account and where all data will e coming from. "n addition, the prolem is split
into different su prolems. #lso the data+set is split into logical partitions such as physical regions, years, etc.Finally, in a preliminary processing step some of the huge data+sets are correlated to e!isting models. These
models are not directly in the main scope of the investigation, ut rather serve as input for uilding the gloalusiness case. =!amples of this are customer adoption and price+evolution of equipment.
"n this phase also the planning horiJon and level of detail will e fi!ed5. There is a trade off etween the
planning horiJon and the confidence level of the model. # longer planning horiJon will e much moresusceptile to &accumulated( errors and as such lead to less reliale results.& 5 This considers the case without cyclical refinement. "t is possile to change the level of detail and planning horiJonetween different cyclical refinements. (
2.2.1.a Cather "nput "nformation
#cquiring data and uilding knowledge of the prolem scope is a difficult and tedious task. Different sources atdisposal for our research areG
K Discussions with equipment vendors and telecom operators within different =uropean, national and ilateral
proLects.
K "nformation found in literature from various authors, generally in the scope of techno+economic research.
K @arious information sources pulicly availale &e.g. FTT- $ouncil 2.2*4, "D#T= 2.284, ?oint Topic 2.254,product search, etc.(
2.2.1. /udividing the ?rolem
9hile the initial research question is fairly small, the prolem quickly increases in oth siJe and comple!ity when
gathering input information. $learly the more information availale, the more realistic the prolem is represented and
the more reliale the optimal solution will reflect the actual optimum. #t this point, Lust after gathering input
information, it is est to structure and aggregate all input information.
#t this stage a high level overview of the prolem is sketched, for instance using the approach as suggested in2.2)4. Further on this overview is filtered using a heuristic selection and prioritiJation step to limit the search
-
8/13/2019 Bakar Optika Pristupna Mreza
6/101
space and solution approaches to e!plore further. Figure 2.16 gives an overview of the high level cost estimation
for an FTT- rollout as detailed later in this dissertation &$hapter * and 8(.
Figure #shikawa breakdown of the cost esti"ate of an FTTH rollout with focus on the i"$act of the $art on the finalcost %indicated b text si&e and weight'
0e!t to an aggregation and limitation of the search and solution space, also the data could e aggregated and
filtered. The main reasons ehind such astractions are the reductions in comple!ity of understanding,implementing and running the calculations. There e!ist a lot of techniques for aggregating or filtering datacontained in large information sets. On very large data sets, this is often referred to as clustering &or in e!tension
data+mining( and a good ackground on those techniques is given in 2.234. /uch clustering can for instance eused to split a region into several smaller regions ased on geographical information, to classify different cities
in a limited numer of groups according to their population density, etc. /tandard statistical analysis &e.g.regression analysis( will e used in e!treme cases, where an information set should e represented y one or a
very limited amount of parameters. The choice of parameters is often driven y logical considerations. =!amplesof its use areG calculating the average return per user ?( for a telecom operator ased on information from
all =uropean countries, constructing a &normal( distriution of the costs for specific equipment ased on differentlist+prices, etc.
2.2.1.c ?rocessing "nput "nformation
This is the final step in the preparation of the input information for the usiness case. "n this phase we process
the input information into logical input models for the further calculations. For instance in the case of thisdissertation, all input data on customer preferences and market dynamics is undled into one of the e!isting
customer adoption models as defined y :ass, CompertJ, etc. "n section 5.1.1.a we will detail the choice of inputmodel and how to retrieve the parameters for this model in more detail.
2.).2 2.2.2
-
8/13/2019 Bakar Optika Pristupna Mreza
7/101
"n this dissertation we did not follow the traditional strict split etween $ap=! and Op=!. 9e propose a
classification ased on required level of detailG
1. ?roportional models
2. Driver ased models
*. Dedicated models
a. ?rocess ased models
. Dimensioning models
Details of each of those models can e found in the following sections. Ceneral modelling encompass oth theproportional models and the driver+ased models from this classification. :oth dedicated models are discussed in
the susequent sections. #s the development of those dedicated models is a maLor part of the contriution in thisdissertation, we refer to chapter * and chapter 8 for a more detailed description of oth models. "t is important to
note that a large techno+economic model might comine different types of models from this classification into
one larger &dedicated( model. #s such the comination of all models discussed in chapter * and chapter 8 will
form an integrated cost estimation model for an FTT- deployment.
2.2.2.a Ceneral
-
8/13/2019 Bakar Optika Pristupna Mreza
8/101
& ) For instance modelling all customer specific costs as a driver ased model, logically linked to the numer of customers ispreferred over a proportional model in which this logical link is not e!isting &e.g. proportional to overall cost((.
Figure !!xa"$le of a $rocess using a flowchart based "odelling a$$roach
Once the process is adequately documented, the cost of e!ecuting this process once can e estimated using for instance
activity ased costing :$( 2.2;4. =ach rectangle in the flowchart represents a task assigned to a person or a team.
=ach diamond in the flowchart represents a conditional split in the e!ecution of susequent steps. :y assigning a cost
to the e!ecution of a task, or in e!tension resources such as time and tools consumed when e!ecuting that task, and
proailities to the diverging paths of a split, analytical methods can e used for estimating the process e!ecution cost.
-
8/13/2019 Bakar Optika Pristupna Mreza
9/101
investment. "nvesting large amounts of money in a proLect always assumes to retrieve this money at the end of
the proLect &in the form of profits( and gain some additional return over this investment. This gain is alwayse!pressed as a percentage per year. The gain one e!pects to get when investing in this proLect will e depending
on the risk of the proLect &loss instead of gain( and the siJe of the investment. $onsidering company investments,this minimal gain is defined y the rate that a company is e!pected to pay to finance its assets. $alculation of this
weighted average cost of capital &9#$$( for a company with a comple! capital structure is a laorious e!ercise
and falls out of scope of this dissertation. Typically for fi!ed access operators this gain is somewhere etweenA.17 and 16.)7 according to 2.*24 and up to 11.27 for the :elgian incument &fi!ed and moile operator(according to 2.**4. This e!pected return is then reflected in the results y discounting all cash flows &see &2.1((
using this e!pected gain as discount rate. #gain all previously mentioned investment analysis techniques canmake use of the discounted cash flows. This leads to a discounted payback time &D?:( and a discounted
cumulative of the cash flows. This latter is often referred to as the net present value &0?@(
Finally also the internal rate of return &"( can e calculated. This is defined as the discount rate at which thenet present value is equal to 6. The internal rate of return seems very intuitive in its use any proLect with an
internal rate of return higher than the cost of capital will also have a positive net present value -owever theinternal rate of return has some important drawacks, also highlighted in 2.*84G
K "t assumes that interim positive cash flows are reinvested at the same rates of return of the proLect that
generated them.
-
8/13/2019 Bakar Optika Pristupna Mreza
10/101
the firm. "t can e used in decisions on product placement, outsourcing vs. own development, strategic direction of the
product &aiming at a niche or at a large customer ase(, operational planning. "t also gives valuale input for a /9OT
analysis &strengths, weaknesses, opportunities and threats(, a tool often used in the evaluation of strategic usiness
options.
The field of usiness modelling is road and provides a plethora of alternative visualisations and approaches.
This falls outside the scope of this dissertation and we refer to 2.*34 and 2.*A4 for more information on the
topic.
2.).8 2.2.8 =!tend ?hase
The previous section often concludes a usiness case or a techno+economic research case. The model is developed, all
cash flows are calculated and discounted, and finally the proLect is evaluated and compared to other proLects using net
present value. The cycle can close at this point and each step can e the suLect of refinement. -owever, also the full
techno+economic analysis can e e!tended in different directions.
The current implementation still provides a limited view on the proLect outcome. There is no information considering
uncertainties and risks. 9hat would happen if prices of equipment decrease faster%slower than e!pectedN 9hat if we
get less%more customersN p to what point &in terms of variation( can we still e safe and what are the overall chances
of a positive usiness caseN 9e use sensitivity analysis to provide an answer to these questions.
The current implementation also disregards all possile managerial fle!iility in the proLect. 9hile uncertainties might
alter the usiness case outcome, they can also in some cases e countered y delierate actions. "t might for instance
e optimal to aandon a proLect halfway if customer uptake is much smaller than e!pected, or speedup a proLect in case
customer uptake is etter. 9e use real option valuation to estimate the positive effect of these actions on the usiness
case.
Finally, the current implementation disregards the roader market picture in which competition and cooperationwill play an important role. This is very important as the telecom market is very competitive. Came theory
provides a conte!t and tools for descriing this competition and evaluating the most likely outcomes of a specificcase.
2.2.8.a /ensitivity #nalysis
9e use sensitivity analysis when we want more information on the possile variations of outcome for variations
in the input. "t gives a roader view on the risks of the proLect. =specially in case we are very uncertain aoutsome input parameters, sensitivity analysis is required.
"n terms of approach, we distinguish etween asic sensitivity analysis and gloal sensitivity analysis. "n asic sensitivityanalysis, we investigate the impact on the outcome of varying one input parameter at a time &keeping the other parametersfi!ed(. The resulting sensitivity information is the variance of the outcome for the given variation of the input+parameter.
Once e!ecuted for all input+parameters, a normaliJed variance can e calculated for each parameter y dividing its ownvariance y the total variance &sum of the variances of all parameters(. This method is optimally suited for a first
investigation as it requires very little computational resources. #ccording to 2.*;4, this approach is not advisale for detailedanalysis, ut rather for the reduction of the numer of input+parameters to take into account in a gloal sensitivity analysis.
"n a gloal sensitivity analysis the different key input+parameters are varied according to a predefined proaility
density function &?DF(, for instance y means of a Caussian, triangular or uniform distriution. $learly thechoice of proaility density function and range over which each parameter will e varied &e.g. standard
deviation in case of a Caussian distriution( will e very important. 0e!t, a
-
8/13/2019 Bakar Optika Pristupna Mreza
11/101
the usiness case, it alleviates &partly the estimation of the risk y means of the discount factor as in the
calculation of the net present value.The approach applied to technical proLects entails the following three stepsG
1. identify the key uncertainties
2. identify the options
*. valuation of the options considering the uncertainties
/ensitivity analysis, discussed in the previous section, is optimally suited for the detection of the key
uncertainties and the same proaility density functions can e used in the real option valuation. The 3+/framework, as proposed in 2.814, can e used as a means for detecting the valid options availale in a given
proLect from a given set of seven option types
Figure "T$es of real o$tions( the )S fra"ework
The value of a proLect should now e e!tended y the value of the options, and is defined as the summation ofthe original 0?@ of the proLect with the value of each of the options.
Figure 2.18 gives a simplified e!ample of an uncertainty tree and the decisions to take, in the second phase of theproLect when rolling out an FTT- network. The decisions on the right hand side will try to decrease the chance
and siJe of losses, i.e. y decreasing or stopping the rollout and possily y selling the network. They will also
try to increase the chance and siJe of profits y increasing the rollout speed when the market is favourale. #ssuch the outcome of the usiness case taking real option valuation into account will reflect the gain of
incorporating fle!iility in the second phase of the proLect.
Figure 1#!xe"$lar uncertaint and decision tree for real o$tions in an FTTH rollout case. Decisions are indicated%right' b a dark background and white text
/everal option valuation techniques are in use in economic literature. :lack and /holes and :inomial treevaluation &and e!tensions( are used in theoretical real option studies. These valuation techniques are generally
-
8/13/2019 Bakar Optika Pristupna Mreza
12/101
not feasile for larger studies in which several options are modelled and several uncertainties are taken into
account. "n this case a
-
8/13/2019 Bakar Optika Pristupna Mreza
13/101
The remainder of this dissertation will proceed on the development of a cost model for the deployment of an
FTT- network. This cost model spans oth the costs for infrastructure and operations of the network, and othare modelled using dedicated modelling approaches. $hapter * shows how dimensioning is est suited for the
calculation of the infrastructure cost and chapter 8 uses operational flowchart ased models for the calculationand optimiJation of the costs for operating the network. This cost model is emedded in the full economic
methodology in two studies in chapter 5. The first study will comine oth cost models and e!plore the tradeoffs
etween oth. The comined cost model is further enriched with an e!tensive design and evaluation step. Thesecond study will make use of game theory and sensitivity analysis to e!tend the research further. $hapter )shows where the same techno+economic ackground has een used in a roader conte!t to evaluate cases in
wireless access, and in the metro and core network. "t also shows where more of the economic methodology wasused. Finally chapter 3 concludes this dissertation with a short overview of the main findings.
%e&erences
2.14 . @an der :erg, H-ow the 0et 9orksG an "ntroduction to ?eering and TransitI, #rs Technica, /ept2, 266A, httpG%%arstechnica.com%old%content%266A%6;%peering+and+transit.ars%8
2.24 "T+T ecommendation >.266G HData 0etworks and Open /ystems $ommunicationsG Open
/ystems "nterconnection+model and notation,I Euly 1;;8.
2.*4 @. $erf, . ahnG H# protocol for packet network interconnection,I "=== Transactions on
$ommunications,
-
8/13/2019 Bakar Optika Pristupna Mreza
14/101
2.1A4 ClasveJelnet #msterdam &dutch( , httpG%%www.glasveJelamsterdam.nl% gnaQalgemeen%inde!.phpN
pageQidS12
2.1;4 $isco, H-ong ong :roadand Driving 9orldRs Fastest esidential :roadand /ervice with $isco?owered 0etwork o/+certified
-
8/13/2019 Bakar Optika Pristupna Mreza
15/101
customers will ecome very important, and the gateway might e an important opportunity for winning and
holding the customer.The rollout of an FTT- network and the complete replacement of all equipment and cales oviously is a very cost+ andtime+intensive proLect. # lot of parameters will have an important impact on the costs ; and the outcome of the usiness case.
# very important choice here is which technology and architecture to use for the FTT- network. 9hile optical networks arealready e!isting for some time in core networks, its application to the access network can still e considered quite recent *.14
with first rollouts in 1;A) to usiness users and 1;;3 towards residential customers. There is still a lot of evolution in
research, standards and equipment. "n the ne!t section we will discuss in detail the different e!isting FTT- flavours and alsogive a glimpse on possile future developments and the impact on the FTT- network rolled out.
Once the technical ackground is set, a detailed cost model for FTT- is constructed. To this aim, we split the network
in three parts outside plant, inside plant and customer premises according to their impact on the cost and on the
calculation approach. For each of the parts we indicate its structure, components and impact on the costs and finally
descrie the models used for estimating the cost &either dedicated or general(. #dditionally, we point where
optimiJations are possile and either show how this can e done, or relate to e!isting prolem and solution
descriptions.
Finally the dimensioning models can e comined with the prices of the equipment to get an overall estimate of
the total infrastructure cost. "n the last su section we show how to do this, point to models for forecasting theevolution of component prices and give indicative values for the prices of equipment.
&A
"t is not uncommon to offer a free use%lease of the gateway, while the actual gateway remains property of the operator.
#dditionally this moves the handoff etween operator and customer up to this gateway, and offers the operator future
opportunities of remote management.;
/ome choices of the operator will also have an impact on possile revenues (
3.1 Fi)re to the *ome"n a fire to the home network, all data is transported using optical signals over a pure glass medium, the fire. Optical fireis installed rather than the other media &e.g. twisted copper or coa!ial( as it has superior physical characteristics. The most
important characteristics of a medium when used for data transport are its attenuation and dispersion16. #ttenuation refers tothe gradual loss in intensity of the signal, or more in general a flu!, through the medium. Dispersion refers to the effect where
different wavelengths, for chromatic dispersion, and polariJation, for polariJation mode dispersion, do not propagate throughthe fire at e!actly the same velocity.
The effects of attenuation increase with distance and might lead over longer distance to a situation in which the
detector can no longer recover the signal. The attenuation of a medium thus determines the ma!imum distanceover which the signal can travel without the use of active amplifiers. The attenuation is also dependent on the
transmission wavelength and Figure *.1 shows the attenuation spectrum of an optical fire, with the three mostinteresting windows indicated in grey. The first window A66nm+;66nm is especially interesting as there are
very cheap silicon ased sources and detectors availale in this window. $onsidering the high attenuation anddispersion only very limited distances are reachale in glass optical fire and it is unusale in FTT-. "t is used
for short distance communication in data+centres and for optical in+house installations using polymer opticalfire &?OF(. The second window 12)6nm+1*)6nm comines a very low attenuation with nearly Jero
chromatic dispersion around 1*16nm &see also elow(. The third window 18*6n+15A6nm has the lowest
attenuation in the spectrum. :oth the second and third window are used in FTT-.& 16 "n addition also noise and immunity to electromagnetic interference will have an impact on the fitness of the medium fordata transport. (
The effects of dispersion are caused y the fact that no laser can generate light of e!actly one wavelength and
polariJation. The effects again increase with the distance and lead to the signal eing smeared open and
interfering with neighouring pulses on the fire 11. The effects of dispersion can e corrected y means ofregeneration, which typically involves a transformation from optical to electrical followed y the correctiveactions and a retransformation to the optical signal. Dispersion typically starts to cause prolems at a range
larger than 16km even when working with low+cost sources, and much larger when working at a wavelength of1*16nm in which there is almost no chromatic dispersion.
&11 The interference etween neighouring pulses will ecome worse when the itrate of the data is increased. (
The distance of medium over which a given andwidth can e transmitted and received correctly considering themediums attenuation and installed equipment, is calculated y means of a link udget as given in &*.1( ased on
*.24. "t is also used to calculate the power of the sources and sensitivity of the receivers required. Tale *.1contains reference values to use for the calculation of the total optical loss for an FTT- network. "t also contains
the link udget otained with equipment according to the standardiJed classes &see also *.1.2(. 0ote that there isalready equipment e!ceeding the highest standards with a link udget of *2d:, referred to as class $B. The
standard values for optical equipment such as connectors and outdoor caling are ased on *.*4. The realistic
-
8/13/2019 Bakar Optika Pristupna Mreza
16/101
values are otained from product information sheets from currently e!isting cale, splicers and connector
manufacturers or vendors &e.g. $orning, *
-
8/13/2019 Bakar Optika Pristupna Mreza
17/101
customers to one fire and one equipment port in the $O. The andwidth of this fire is shared amongst the differentcustomers. $urrently time division multiple!ing &TD
-
8/13/2019 Bakar Optika Pristupna Mreza
18/101
*.1.2.a :?O0, C?O0 and 0C?O0
"n 1;;5 the full service access network &F/#0( group was estalished as a consortium of operators. They provided the
first FTT- ?O0 standard initially called #?O0 and later on re+randed to :?O0 standardised from 1;;A onwards
under the umrella of "T+T in the C.;A*.1 through C.;A*.5 &*.84+*.A4(. "t initially selected #T< as their layer 2
signalling protocol which was roadened to other layer 2 protocols in the course of standardisation. Original #?O0
offered a 155.52 16 and 1666:ase+?>26 with a ma!imal reach of respectively 16km and 26km. # typical
split ratio of 1G1) is defined and more is possile. 0e!t generation =?O0 or 16C=?O0, egan standardiJation in266) under A62.*av working group *.134 and will offer a symmetrical andwidth of 16Cps and a ma!imal
reach of 26km. Downstream and upstream traffic is usually sent on the 18;6nm &or in e!tension 1556nm(wavelength and the 1*16nm wavelength, respectively, ut this is no restriction. Finally the 1556nm wavelength
can also e used in =?O0 to overlay a third channel of communication.
*.1.2.c #O0 /tandards
#s mentioned efore, the =F< already envisaged point+to+point =thernet communication for FTT-. The specifications
however are in line with the ?O0 specifications and as such disregard the specific setting of a dedicated fire connection."=== A62.*ah defines 1666:ase+:> and '> for single and doule fire point+to+point connections at a nominal rate of
1Cps full duple!. #ll wavelengths can e used for the transmission of data typically '> will prefer two times 1*16nmwhile :> will use 18;6nm in downstream and 1*16nm in upstream. The standards define a distance up to 16km Lust like for
:?O0 where in #O0 a much higher distance can easily e reached due to the much lower optical loss &no splitters(.$onsidering a link udget of 26d:, comparale to the lowest defined equipment class in ?O0 networks, this could reach adistance up to )6km.
9ithin "T+T C.;A5 *.1A4, the specification of F/#0 for point+to+point FTT- communication is given. "t
descries a andwidth of 166
-
8/13/2019 Bakar Optika Pristupna Mreza
19/101
significantly. The working groups "=== A62.*av, "=== A62.*a and "T+T C.ge are e!amples of this move
in standardisation.
K'anufacturing and installation practices. -igh+end optical equipment has een historically used e!clusively inthe core network. #s such its deployment and maintenance was restricted to more skilled personnel using
e!pensive automated tooling. "n the access network, prices for the installation will take a larger part of the
costs per customer and reducing this cost at the e!pense of some core+grade requirements &e.g. less finesplice quality( can lead to large overall cost savings. #lso equipment manufactured with techniques notqualified for use in core networks, might e still fit for use in access, especially when this would lead to a
much lower production cost.
K &ntroduction of wavelength division multiple#ing. 9D< allows multiple wavelengths to e transmitted overone fire. "n the case of core networks dense 9D< &D9D
-
8/13/2019 Bakar Optika Pristupna Mreza
20/101
additional functionality &e.g. switching%routing, video server location, data+server, authentication, etc( can e provided.
# /$ is much smaller, connecting *2+512 customers. "t is placed in the street and requires electricity and cooling. "t
contains the same, or comparale equipment as in the $O in case of an -0 and its costs can e modelled in the same
manner. $onsidering the limited space in the /$, the remaining active equipment will often still e placed in the $O.
The outside plant connects the inside plant to the customer. The outside plant in this astraction is restricted to passive
equipment and stops where active equipment is placed. "n the case of an #/0, the outside plant encircles the active
islands as the street cainets are considered inside plant from a cost modelling perspective. The outside plant contains
all passive aggregation points ?( and fle!iility points &F?(.
#t the customer premises, the outside plant is restricted to the entrance of the house. :eyond this point, we
distinguish the customer installation &also active(.
Figure 14High level overview of an FTTH network fro" a cost "odelling $ers$ective with a distinction between +S,and H, /O,
"n the following su sections we descrie these three parts in more detail and uild a cost model for each of
them. The order in which we tackle them, outside plant, inside plant and customer premises, is ased on theircost, taking the largest cost first.
*.2.1 Outside ?lantThe outside plant consists of all passive FTT- equipment. This contains all ducts, su ducts and fires. "t also
contains all passive infrastructure intended for maintenance and repair purposes and for incorporating fle!iilityor future e!tensions. This equipment consists of all types of closures such as hand holes, man holes, street
cainets and pedestals.
Figure 15Sche"atic view on the installation and $hsical cable infrastructure in a buried outside $lant.
-
8/13/2019 Bakar Optika Pristupna Mreza
21/101
"n the case of a new FTT- network rollout, all the equipment will e installed in the streets &e.g. digging trenches,
fi!ing cales, etc.(. This often represents a very high part of the work and cost, and in case of a fully uried outsideplant the cost of trenching even dominates all other costs and consumes somewhere around 367 of the total cost of the
deployment &see results in 5.1.*.a(. $onsidering the siJe of this cost, it is clearly a focus of optimiJation. Figure *.8shows the situation of a fully uried installation.
Finally the outside plant also contains all passive equipment installed in the FTT- network. 0owadays thisincludes all passive optical splitters installed in a ?O0. #s mentioned efore, in the future this might as well
include passive wavelength splitting equipment such as an #9C.
*.2.1.a "nstallation
# very high cost is caused y the installation of the outside plant. "n case of a fully uried outside plant thiscomes down to digging trenches and installing ducts in the trenches. Typically these ducts would already contain
either fire cale or micro+ducts. The cost of digging trenches will e very dependent on the soil andunderground situation and some mean values are indicated in Tale *.8 at the end of this chapter. The same tale
also contains information on the cost of a duct and fire cale, and this clearly shows why trenching willdominate all other outside plant costs &and often even inside plant costs(.
Different analytical models can e used for estimating the installation length. #ppendi! # gives an overview of differentanalytical installation topologies when considering customers which are uniformly spread over a given area. "n all models we
indicate the distance etween two neighouring houses y l and the length of one siJe of the square &in numer of houses( yn. "n all situations the central office is placed in the middle of the square, as this will reduce the length of fire to e installedin the trenches. "n all uried installations, crossing a street will cost much more &per meter( than all other trenches installed
along the streets. "n case of aerial deployment, or when e!isting ducts can e reused this difference is small or non+e!istent."n the analytical models taking this difference into account, we indicate the street+width y w. 9e differentiate etween the
analytical models according to the type of installation they resemle the most all uried or aerial. #ll analytical modelsreturn an installation length function #.n2.l, in which # is a factor depending on the installation type. as there are n2
customers in the area, a dedicated length of #.l must e installed per customer. Depending on the analytical model, the factor# is etween V6.; &simplified /teiner tree( and V1.5 &doule street length(.
The different analytical models can e used to estimate the installation length ased on the customer ase and
area siJe. The choice of the most appropriate analytical model is ased on the prolem at hand. -owever,considering the dominant siJe of the trenching costs for a fully uried FTT- deployment, a etter dimensioning
is often still required. The full optimal trenching topology can e calculated y constructing a /teiner+treeconnecting all customers. The /teiner tree solves the prolem of connecting 0 points with a minimal cost tree
structure *.2)4 &p. **;( with the freedom to use any numer of additional intermediate points. The /teiner treeover a uniformly distriuted customer ase is shown in the last analytical model. "n a realistic setting with
e!isting streets and houses, we consider the /teiner tree prolem in graphs. The formal definition of this prolemisG Civen a weighted graph &/,0,w(15 and a vertices suset / @, find a tree of minimal weight which includesall vertices in . This prolem is 0? hard and we thus relied on e!isting heuristic methods *.254 for calculating
a near optimal tree structure. 9e implemented the #
-
8/13/2019 Bakar Optika Pristupna Mreza
22/101
-euristic algorithms such as proposed y *.254 can e used and an implementation for a ring constructing
algorithm was made.&1) #ll edges in the remaining, non+removed parts of the tree will have a Jero cost associated as they are already scheduled toe opened for installation. The new connected tree might use some of those edges to reconnect and as such this might lead to
a local optimiJation.13 This level of detail is not availale in currently e!isting C"/ ases1A
# -amiltonian cycle is defined as a cycle in an undirected graph which visits each verte! once and also returns to thestarting verte!.1; The vehicle routing prolem is defined as finding the optimal set of routes for a fleet of vehicles in order to serve a given
set of customers. #s such it is related to the -amiltonian cycle prolem *.2)4 &p. *11( in which case the fleet consists of one
vehicle.26 The capacitated vehicle routing prolem is an e!tension of the vehicle routing prolem in which each vehicle has a
ma!imum numer of customers it can serve. (
Figure 16Fa0ade and +erial installation as alternatives for a buried outside $lant
'arge savings are possile if &part of( the network can e installed without trenching, for instance y means of aerial
deployment &see Figure *.5(. Other possile solutions e!ist, such as reusing e!isting ducts &even after e!tracting the e!istingcopper cales *.2;4(, using the sewage system *.*64, attaching the cales to the facades, etc. "n case the network has een
rolled out, also the sharing of the installation for different infrastructures &gas, electricity or water( might lead to important
savings. This will especially e the case for a non+telecom infrastructure provider, such as an electrical utility company. "n afull cost model the different installation alternatives must e taken into account in constructing the cost model. The lease offire &or duct( connections to other operators might also lead to important savings in this conte!t &see also chapter )( 21.
Finally an operator will further try to increase chances of profitaility for the network rollout, y focusing on
those areas with the &estimated( lowest installation costs and highest &e!pected( revenues. Figure *.) &left( showshow the costs of connecting only the closest &to the $O( customers will reduce more than linearly. The same
Figure *.) &right( shows how the installation costs evolve for an increasing residential density &according to thesimplified /teiner tree analytical model(.
&21 # telecom operator might try to close his network to other operators as much as possile as this could give him a
ompetitive advantage. "n case of a non+telecom infrastructure owner, such advantage is generally non+e!isting and he will tryto lease as much free infrastructure &ducts, fires, etc.( as possile.(
Figure 1#"$act of focussing on the "ost $rofitable areas first* b not connecting all houses in a given area %left' andb selecting those areas with the highest residential densit %right'
/o called cherry+picking, when performed coarse+grained, will typically e ased on residential density and as
such favor cities over rural areas and the centre of the city over the suurs. =!isting studies in the literatureoften distinguish three or four types of areas rural, semi+uran, uran and dense uran with profitailityincreasing in the same order. The operator will then start picking the dense uran areas first and go down up to
-
8/13/2019 Bakar Optika Pristupna Mreza
23/101
the point at which profitaility is no longer sure. The same cherry picking approach can e performed more fine+
grained. This is often referred to as geo+marketing and uses more input sources 22 as detailed as possile &alsoover much smaller customer ases(. $lustering techniques *.*14 are used for grouping the most relevant
customer ases and get for each an indication of the chance of profitaility. $onsidering the cost of trenching thedistance to connect all customers will e a very important parameter and the previously mentioned analytical
models can e used for a fast estimation of this cost, ased on the area surface and inhaitants.
*.2.1. Ducts and Fires
The outside plant contains all fire strains connecting the customer to the central office, either directly or in apassive or active aggregation structure. sing such aggregation or not will have a maLor impact on the fire
count in the network. For easier comparison of the different architectures, the network is split into two differentparts. /tarting from the central office, the feeder sections runs up to the first split+point. #t this point in an -0,
the high fire count feeder cale is split into several cales containing a lower fire count per cale. "n this pointin a ?O0 splitters are used, while in an #/0 active equipment can e used for the aggregation of the traffic. One
or more distriution sections stretch from this& 22 For instance in the case of FTT- the operator might e interested in following information over the area &ne!t to
residential density of course( susoil, income of inhaitants, e!isting customer ase, age of inhaitants, adoption of previous
technology, T@+set siJe of inhaitants,X (?O0, -0( the last distriution section will always contain the same amount of fires, equal to the amount ofpotential customers.The fire can e easily dimensioned given the topology of the full installation rollout &e.g. analytical models or /teiner tree(.
"t suffices to fi! the aggregation siJe and the numer of aggregation points. unning through the topology from the leaves upto the root allows finding the location of the aggregation points and calculating actual siJes. "t should e noted that the actual
siJe does not need to e equal to the predefined siJe 2*. #ppendi! # contains also the analytical models for calculating thefire length necessary for covering the whole area with one dedicated fire to each customer and ased on the topology
models introduced efore. #s mentioned, an #/0 and ?O0 will significantly reduce the fire count y the insertion ofsplitters in the network. #ll analytical models have een constructed from a hierarchical structure and taking this reductioninto account oils down to a division of the fire count at the right level. Due to the highly repetitive structure of the
analytical models, aggregations of fires y a power of 8 are more easily calculated.
The numer of fires will also determine the numer of splices required to install the outside plant. "n the case of
a ?O0 the fires will e spliced on the splitters at oth sides. "n oth the case of a ?O0 and #O0, all fires wille spliced on connectors at the customer side and at the $O &or /$ in case of #/0(. "n the case of an -0,
several fire cales could e merged onto one cale with a higher fire count in order to reduce the consumedduct space, and cale and fire management comple!ity. The length of a fire cale on a reel &wheel( is not
unlimited and typically a reach of 5km to 16km is possile, eyond which splicing is required. $alculating thenumer of splices required in the network is given y &*.2(. This value could ecome very large in case of a high
fire count in comination with a long distance to the central office.
&2*
#n operator will typically use a limited set of equipment, as this might reduce comple!ity in the installation and improve
uniformity and internal standards for the company. #s such the predefined aggregation siJes will often e a clear upper limit
for the actual aggregation siJe. (
*.2.1.c ?assive =quipment0e!t to the physical installation, fires and ducts, the network contains various types of additional &passive(
equipment such as man holes, hand holes, street cainets, poles, etc. # schematic overview of such passiveequipment is shown in Figure *.A. The required amount of this equipment is more or less the same for all
different architectures and is typically dimensioned using a driver ased approach. -and holes and man holes areinstalled ased on the total topology length, with one hand%man hole more or less each 566m *.*24 &p. 2A(. #
fle!iility point &F?( can e installed at the last distriution section 25. "t adds fle!iility to connect customers
when required. #s such an initial F? can e
& 28 "n some papers &e.g. *.**4 p. )1(, the required duct space is added to the final result as an e!tra opportunity cost. Thereasoning eing that a less occupied duct can still e leased to other operators and as such generate additional income which
is lost when completely filling the duct. 9e did not adopt this approach for two reasons. Often the underground &e!isting freeduct( infrastructure is not sufficiently availale to provide e!isting usale ducts to cover a sustantial part of the network and
a full installation of new ducts is necessary anyway. /econdly, in case a fire is rolled out to each customer, the operator canalso lease this fire or itstream access over this fire to other operators and as such makes profit of the filled duct. "n this
case there is less need for empty space in the duct.
-
8/13/2019 Bakar Optika Pristupna Mreza
24/101
25 "t can easily e installed at other locations as well. 9e use F? when this contains only passive equipment, typically a patch
panel and splitters. 9e use /$ for pointing at active equipment at short distance from the customers as is the case in an #/0.
(
smaller than the potential customer ase and e!tended in case the take rate is higher than the anticipated market
share. "n a ?O0 with so+called centralised splitting, this F? will also contain the splitters
2)
. "n distriutedsplitting, the splitters are installed at different locations and might involve cascaded splitting, for instance whenconstructing a 1G*2+split y placing a 1G8 close to the $O and a 1GA close to the customer. The enclosure at which
the dedicated cale to the customer is connected is often called the drop+o!. "n an #O0 this will split a calecontaining several fires into different cales containing one or two fires each. "n a ?O0 this can additionally
contain a splitter and as such use an incoming cale containing fewer fires.
Figure 1!Sche"atic overview of $assive e1ui$"ent re1uired in the installation of the outside $lant
&2)
#s such a future upgrade to a smaller split+ratio or a 9D< ?O0, y installation of #9C at this location, can e
performed much easier. #lso diversified service offerings, y means of a different split ratio, are possile at this location ut
should not e!ceed the ma!imum difference in optical udget etween two customers in the same ?O0. (
2.).A *.2.2 "nside ?lantFor the evaluation of the costs, we split the inside plant in three large functional locks Fire termination,
active equipment and the connection to the metro network as shown in Figure *.;.
Figure 1"Sche"atic overview of the inside $lant e1ui$"ent
Fires are connected to the equipment y means of standardiJed connectors. #s such, all fires entering thecentral office &or street cainet( will e e!tended with a connector. There are a lot of fires entering the uilding
and in order to keep a good overview of which fire is connected to which equipment, the operators splice thefires to a physical &management( frame, often called the Fire%Optical%
-
8/13/2019 Bakar Optika Pristupna Mreza
25/101
The connectors at the ackside of the ODF will e connected to equipment providing connectivity to the
customer, the so+called optical line termination &O'T(. This is rack mounted equipment, containing a ackplanewhich aggregates all traffic of the incoming fires &front( and directs this to the ack. There are different possile
protocols for oth the traffic to%from the customer &front( and the traffic at the ack. For instance a TD
-
8/13/2019 Bakar Optika Pristupna Mreza
26/101
The list of equipment contains very different types of equipment. Cenerally speaking we distinguish etween
active optical, passive optical, electronic and mechanic &passive( equipment. These types of equipment differ alot in physical lifetime and price evolution.The physical lifetime is often e!pressed y means of the mean time to failure &
-
8/13/2019 Bakar Optika Pristupna Mreza
27/101
?rices for equipment are hard to find and often not very reliale due to the high discounts up to )67 largercompanies can get from the vendors on list prices. Tale *.8 indicates values which have een found in various
pulic sources and which have een checked during meetings and discussions with vendors and operators in thescope of various proLects.
& 2; 9e used the values as indicated in the tale for most of our calculations. They are ased on the values presented in *.*84in 2668. 9e still assumed FTT- &and its deployment( to e in a rather novel production phase in comparison to activeelectronics. "n a later stage we e!pect the parameters of active optics to change to those of active electronics. ?assive optics
will most likely evolve into an even more mature production phase. (
&ndicative67)rices for F equipment for a full deployment*6
#ll prices are indicative values found in pulic sources at the time the study was performed.
-
8/13/2019 Bakar Optika Pristupna Mreza
28/101
#pplying these dimensioning models to a specific setting &area, customer ase, etc.( lead to a list of equipment
&and installation( which in comination with their price leads to the overall e!penditures. The last sectioncompletes the infrastructure dimensioning with a description of commonly used price evolution curves and a
tale with an HindicativeI price of equipment.
%e&erences
*.14 -. /hinohara, H:roadand #ccess in EapanG apidly Crowing FTT-
-
8/13/2019 Bakar Optika Pristupna Mreza
29/101
*.1A4 "T+T ecommendation C.;A5, H166 Technology, httpG%%www.kael+!.com%inde!.php
*.*64 . uderman, HFrance Telecom plans massive FTT- roll+out in 266;I, 'ightwave =urope,Eanuary 266A, httpG%%lw.pennnet.com%displayQarticle%*21*6;%)*%#T$'%none%none%1%France+Telecom+
plans+massive+FTT-+roll+out+in+266;%
*.*14 ?. erkhin, /urvey of clustering Data
-
8/13/2019 Bakar Optika Pristupna Mreza
30/101
A--lica)ilit/ in a Broader Technological or
Economic $onte0t
One $ound of learning re1uires ten $ounds of co""on sense to a$$
)ersian )roverb
)roverb"n this dissertation we developed a methodology for the evaluation of investment proLects intelecommunication networks. The main focus was on the evaluation of the usiness case for an FTT- deployment, and
a lot of work was devoted to the development of infrastructure and operations dimensioning models. The
methodology and the specific models developed in this dissertation can also e applied in other conte!ts. "n our
research, we have used the same approach outside the scope of an FTT- deployment study. "n the first sectionwe roaden the technological scope of this dissertation. "n section ).1.1 we e!tend the work to include fi!ed
access networks. 9e further open this in section ).1.2 to wireless access network deployments. Finally weshortly show the limited applicaility in metro and core networks §ion ).1.*(. 9ithin the dissertation, we
e!plored the evaluation of the FTT- deployment with the e!tension of game theory and sensitivity analysis. Themethodology encompasses other economic studies and e!tensions not covered efore. -owever as some parts of
our research also considered those e!tensions, we discuss the approaches and results in section ).2.
6.1 E0tending the Technological tudies
The techno+economic study in this dissertation was focussed on the fi!ed access network. "n chapter 2, we
looked at the network from a much higher point of view and gradually refined the description to get up to thepoint of the FTT- access network. "n this section we gradually Joom out again. #t each point, we indicate which
models could e easily reused. 9here applicale, we add results from our e!isting studies.
6.1.1 From FTT* to Fi0ed Access Networks There are good resemlences etween FTT- and otherfi!ed access networks, and as such most of the information discussed in this dissertation can e reused there.
$sidering the infrastructure, a large part of the calculations in the FTT- case were devoted to the dimensioningof the outside plant installations. This calculation can e fully reused when considering a new installation of
other fi!ed access technologies &e.g. in a newly uilt neighourhood(. #lso when looking at upgrades of e!istinginfrastructure, e.g. from #D/' to @D/', additional fire will e installed. "n this case a lot less optical nodes
&e.g. @D/' street cainets( should e connected with fire and a full optimal /teiner tree can e more easilyconstructed. Often an operator will connect different optical nodes in a ring. Finally it is also very
straightforward to translate the calculation of all other costs such as inside plant equipment and customerpremises equipment to a non+FTT- situation. 0e!t to infrastructure, also large parts of the operational modelling
can e reused in the scope of the deployment or e!ploitation of other networks. For instance, we can e!pect moreor less the same operations for keeping the network up and running. "n the case of an e!isting network
infrastructure, the physical installation of the customer connection process, i.e. connecting the cale to thenetwork, will not e triggered a lot as most customers are already connected to the network. "n order to reduce
additional travelling to%from the customer, operators try to promote a do+it+yourself installation 53 as much as
possile. "n chapter 8, we have already shown the differences etween a fire ased and a copper ased accessnetwork. "n ).14 we made a cost comparison of the repair process for oth network technologies. Figure ).1
gives an overview of the costs for one occurrence of the repair process in a copper ased network &according toFigure 8.15(, a preventive repair for copper, i.e. a solution for the effects of water intrusion &also shown in Figure8.15( and the repair in a fire ased network &according to Figure 8.18(. #ll assumptions for this study case can
e found in ).14. $learly the high amount of fires in the feeder section of an -0 will
&53 This goal to reduce all additional travelling to the customer is also an important driverfor remote management.(
cause the repair cost for a failure in this part to rise significantly5A. "n the distriution section this is less clear. Fin
architecture. till there is a much lower distance trenching
feeder section than in the customer connection and distriution section. The feeder section often has a length
smaller than 2km and is shared over a large customer ase. "n this study case 2666 customers were connected to
-
8/13/2019 Bakar Optika Pristupna Mreza
31/101
one feeder cale over a length of 1666m &uran( up to 1566m &rural(. $ustomer connection often requires a
dedicated cale with a considerale length per customer &here A8m &uran( up to 1)Am &rural( was used(. Thisleads to a relatively low impact of the feeder section in the overall yearly repair costs. "n all calculations we
assumed the occurrences of cale cut in copper &reactive repair( to e equal to those in fire. 9e also assumed *times as much preventive repair actions in the copper network as reactive repair actions. The results indicated in
Figure ).2 show how a ?O0 has the lowest repair costs, followed y the copper network. #n -0 has
sustantially more fires in the network and will clearly have a higher repair cost up to **7 higher &uran( thanin case of a ?O0. ally in the customer connection section, the same amount of fires or copper pairs are used percale regardless of the traLectory( in the
Figure 212ost $er occurrence of the re$air $rocess for co$$er reactive$reventive and FTTH H, /O,
&5A The peak of -0, which is as high as 26; cost units, is left out of the figure and replaced y an arrow and its siJe &for thesake of clarity.(
Figure 22Overall ti"ated re$air cost for co$$er and FTTH H, /O,
8.62onclusionsThe methodology and models developed in the dissertation are focussed on the deployment of an FTT- network,
ut they can easily e used in a roader conte!t. 9e have first shown how most of the same models and
information could e used in the scope of other fi!ed access networks &e.g. D/' or -F$(. "n a second step wehave shown how large parts, especially of the operational modelling, can still e used in a wireless access
-
8/13/2019 Bakar Optika Pristupna Mreza
32/101
network deployment. 9e related in oth cases to results from techno+economic studies we performed. The
economic methodology presented in this dissertation can also e easily roadened. egarding the inputmodelling of the usiness case, we have investigated how the pricing could influence the customer adoption and
as such the whole usiness case. "n a second e!tension, a full usiness model of the FTT- deployment isconstructed that sheds a light on the choices and opportunities for opening the FTT- network to other operators.
Finally the theory of real options, and more in detail the 3+/ framework, has een applied on the case of an
FTT- deployment. The different options found here could easily e used to e!tend the strategy set of the gametheoretic evaluation used in chapter 5.
WIK-Consult ReportStudy for theEuropean Competitive Telecommunication ssociation !ECT"
The Economics of #e$t%eneration ccess -
&inal Reportuthors'(ieter Eli$mann(ra)an Ilic(r* Karl-+ein, #eumann(r* Thomas l.c/e0aumWIK-Consult %m0+Rh1ndorfer Str* 2345267 8ad +onnef%ermany
8ad +onnef9 Septem0er :69 ;663
=!ecutive /ummary1. The =uropean $ompetitive Telecommunications #ssociation &=$T#(
commissioned this study on YThe =conomics of 0e!t Ceneration #ccessY from9"+$onsult on #pril 2;. The main oLective of the study is the assessment of the
viaility of ne!t generation access usiness models and the analysis how regulationmight support viale duplication of infrastructure whilst ensuring competition in the
provision of services to consumers and usinesses. The study also gives anoverview of recent studies dealing with 0C# in =urope and aroad ustralia,
/ingapore, Eapan and the /#(.
2. To meet the oLectives of the study, we have developed a generic usiness model.This model on the one hand enales the assessment of the viaility of ne!t
generation access usiness models and the potential national coverage of 0C#.On the other hand it provides the opportunity to derive conditions &in particularregulated wholesale services( that allow a ma!imum degree of viale duplication
&replicaility( of a first movers investment. -owever, as the term HgenericIsuggests, the model structure and logic itself is independent from the conditions in
a specific country or area. ather, the model requires to feed in real world dataaout a country or an area within a country in order to generate actual empirical
results &which are then country%area specific(. The model for the first time givesparticular emphasis on the impact of certain regulatory decisions on access
regarding 0C#.*. There are already several studies that focus on various facets of fire deployment
in the access network. 'ikewise, there are already several models availale whichare focusing on certain aspects of the viaility of deep fire deployment. The
present study covers all relevant aspects in a comprehensive way. 9e review therelevant literature and models, we analyse actual fire deployment approaches in
-
8/13/2019 Bakar Optika Pristupna Mreza
33/101
several countries in and outside =urope, we present the main features of our model
and we apply this model to generate empirical results for altogether si! =uropeancountries &Cermany, France, "taly, ?ortugal, /pain, /weden(. For all si! countries
we provide comprehensive empirical evidence on the viaility of replication of@D/'%FTT$ infrastructure as well as of the deployment of FTT:%- infrastructure.
9e also show quantitatively and not only qualitatively the impact of regulatory
measures like duct and dark fire access, fire loop and su+loop unundling onthe replicaility of 0C# roll+out and competition. On the asis of our model resultswe derive recommendations on the necessary regulatory conditions for effective
competition in 0C#.8. 9e have structured the model to calculate fire deployment for eight coverage
areas or HclustersI in each country defined y population density with the
e!pectation that profitaility of 0C# deployment depends on population density4* We assume anadvanced state of net
-
8/13/2019 Bakar Optika Pristupna Mreza
34/101
re?uirements for an operator
-
8/13/2019 Bakar Optika Pristupna Mreza
35/101
de)ree of replica0ility* Where via0le9 replication of the incum0ents #% re?uiresamore si)nificant scale andAor mar/et share for alternative operators compared
-
8/13/2019 Bakar Optika Pristupna Mreza
36/101
due to their si,e and ris/ position*:2*4 (ue to the factors mentioned a0ove9 investments in #% are more ris/y foralternative operators than for incum0ents* Let9 alternative operators may actas first movers in #% 0ecause their current 0usiness model as a
-
8/13/2019 Bakar Optika Pristupna Mreza
37/101
0e realistically achieva0le in a competitive environment* &i0re G and SG arealso the prere?uisite for )ettin) !at least" the same de)ree of competition asunderthe current un0undlin) model in the ST#*;:* The follo
-
8/13/2019 Bakar Optika Pristupna Mreza
38/101
structural issue relatin) to the hi)h costs of layin) physical infrastructure in atypicalEuropean environment !medium density9 0uried ca0les"* olicy-ma/ers should setrealistic o0ectives for re)ulators on this 0asis
-
8/13/2019 Bakar Optika Pristupna Mreza
39/101
57* Certain models of ris/ re
-
8/13/2019 Bakar Optika Pristupna Mreza
40/101
means in particular that relevant access products are not only availa0le inprinciple0ut are effectively availa0le in due time*5* Re)ulators should do more than they did
-
8/13/2019 Bakar Optika Pristupna Mreza
41/101
deployment in the access net
-
8/13/2019 Bakar Optika Pristupna Mreza
42/101
then countryAarea specific"*There are already several studies that focus on various facets of fi0re deploymentin theaccess net
-
8/13/2019 Bakar Optika Pristupna Mreza
43/101
specific approaches and e$periences or)an !;662"9 DEC( !;663a9 0"*;*: DT' 8usiness cases for 0road0and access;*:*: DT' 8usiness case for su0-loop un0undlin) in the #etherlandsDn Fanuary ;79 ;66@ DT fundamentally revised its proposed previous positionon
K#s ll-I proect follo
-
8/13/2019 Bakar Optika Pristupna Mreza
44/101
areas served9 it may 0e economically via0le to deploy SG to around :9666 ofthe lar)est street ca0inets in dense ur0an areas !out of a total of around ;39666street ca0inets in the #etherlands"9 provided that'o the interconnect and
-
8/13/2019 Bakar Optika Pristupna Mreza
45/101
per month* K# can maintain a 26= mar/et share of the total 0road0and customer 0ase!74= retail customers9 :4=
-
8/13/2019 Bakar Optika Pristupna Mreza
46/101
of 56 per metre for duct costs9 the monthly cost of the un0undled fi0re loopamountsto :@* * This fi)ure is the /ey input fi)ure for the alternative operators0usinesscase* &i)ure : represents the results for five different )eotypes or clusters* Theyareordered 0y line density9 a similar approach as
-
8/13/2019 Bakar Optika Pristupna Mreza
47/101
>0ps" in K#s current reference offer* Dnly if these prices ar/et share9 si,e of ca0inets tar)eted9 additional revenue from SG9 0road0and penetration9
costs ofeircomJs
-
8/13/2019 Bakar Optika Pristupna Mreza
48/101
Figure 28vera)e cost per line - 0ase and optimistic scenario for SG compared(&* The study concludes that it is unli/ely thatcompetitors
-
8/13/2019 Bakar Optika Pristupna Mreza
49/101
Ensure sufficient collocation space in eircoms street ca0inetsM Ensure availa0ility of an afforda0le fi0re 0ased 0ac/haul product from eircom*
+' #nalss- ibre in the 2ast Milenalysys !;662" has constructed a simple )eneric model of RDI for an operator ina socalledtypical Western European mar/et of :6 million households*
AssumptionsIn this model it is assumed that the incum0ent 0uilds out ne< access net
-
8/13/2019 Bakar Optika Pristupna Mreza
50/101
mar/et and a more competitive TB mar/et* &inally9 Scenario ( is characteri,ed 0yareceptive consumer mar/et and a more competitive TB mar/et*
The 0est-case scenario for a telco !scenario " is characteri,ed 0y a stron) ta/e-up ofmulti-play services across the operators user 0ase* +o
-
8/13/2019 Bakar Optika Pristupna Mreza
51/101
-
8/13/2019 Bakar Optika Pristupna Mreza
52/101
Figure 26 7I on V/82 deploment
2.5 Aisem studies &or A%$E
The French regulatory authority #$=? has identified three maLor issues for FTT-
deployment which areG &1( access to France Telecoms civil engineering facilities, &2(
sharing of the terminal part of FTT- and &*( intervention of local authorities as
facilitators. #$=? has commissioned studies from #visem for the second and third
issue. These studies are focused on uran and suuran regions of France where it ismore likely that FTT- deployment will take place. "t addresses technical feasiility of
FTT- deployment while it does not consider legal issues such as property rights or
restrictions on rights of way.
2.5.1 /haring of the terminal part of FTT-
/cope
#visem &2663a( considers how e!isting infrastructure could facilitate the entry into
uildings for FTT-%FTT:. "t descries the technical options for getting inside a uilding
via eight different infrastructures and it evaluates the feasiility of leveraging them to
install fire access to the uilding. The infrastructures considered are those for supply of
electricity, France Telecom infrastructure, cale T@, waste and fresh water
infrastructure, gas, community heating, and street lighting. The study elaorates on fivedifferent methods of access to the uilding which are underground, overhead, faZade,
mi! of overhead%underground and mi! of underground%faZade. "n addition it
differentiates etween single family homes and multi+dwelling units. The study focuses
on fundamental technical feasiility considerations and suggests measures to increase
the aility for leveraging the analysed infrastructures. -owever, the study does not
directly address the economics of utiliJing each of the infrastructures.18
esults
K tiliJing ducts and other facilities of France Telecoms network ears high
potential. The study addresses duct sharing, underground chamers and further
assets such as pole infrastructure. -owever, at the time of completion of the
study, a commercial offer e!isted only for duct access. Otherwise stated, nooffer was availale regarding access to overhead infrastructures such as poles,
-
8/13/2019 Bakar Optika Pristupna Mreza
53/101
to manholes independent of ducts and other elements that could e leveraged.
K =lements of $#T@ networks could also ecome relevant assets. -owever,
according to the study there has een little agreement with France Telecom
&FT(15 on using the $#T@ ducts &which are in many cases owned y France
Telecom( to get access to the suscrier and even less for ducts etween
manholes, due to FTs e!clusive usage rights clause in the agreements with the$#T@ companies.
K =lectricity networks could potentially e leveraged y sharing the use of poles
&possile ut with numerous limitations, e.g. currently only 2 operators may
share pole infrastructure at a time(, y sharing ducts for remote energy
management &unlikely due to different prolems( and faZade mounting &as it is
done y a $#T@ susidiary of =lectricitP de France(.
&18 The only element of information is provided in the form of selected prices for accessing e!isting
infrastructure &e.g. France Telecoms current duct offer or pricing for =lectricitP de Frances overheadelectricity infrastructure for deploying $#T@(.
15 The study details that 2%* of French underground $#T@ networks have een deployed y France
Telecom, using almost e!clusively France Telecoms e!isting assets without the need for further civilworks. Only 1%* have een deployed y independent operators requiring new civil works.(
K FTT- deployment via sewage networks is technically possile oth in
accessile and non+accessile underground drains. "n the latter case roots
may aid in deploying cales. -owever, generally installations must address a
numer of potential issues &such as damages from floating material,
deterioration, rodents, pressureX(.
K "n addition to the requirements of sewage networks, FTT- deployment in fresh
water networks requires &among other issues( to circumvent valves, to respect
specific deployment rules &e.g. regarding cale material( and to &*( take into
account the interruption of water service.K =!ploitation of gas networks can e possile ut must respect similar constraints
as deployment in fresh water environment &vents, security(.
K "t appears difficult to use district heating networks for FTT- deployment due to
the inherent high temperatures.
K /treet lighting does not offer many assets as it does not provide uilding entry
e!cept for underground%overhead transitions to the uilding front.
The report suggests a numer of measures to increase the potential offered y the
different infrastructures. =!amples includeG
K egarding &overhead( energy network infrastructure &=DF(G 0o a priori and
systematic limitation of the numer of operators.1)
K egarding France Telecom or $#T@ infrastructureG #llow sharing of all relevant
infrastructures.13 ?rovide municipalities with the same rights to access
infrastructure as a telecommunications network operator.
K egarding faZade entry to uildingsG /tudy technical and legal feasiility to
deploy fire attached to e!isting caling taking account of the specific types of
cales &electricity, telephone, cale T@(.
K egarding water and gasG =ncourage the identification of plants that are not in
use anymore and which could facilitate FTT- deployment.
&1) The current guide regarding electricity networks &Hguide pratique des appuis communsI( limits thenumer of operators which can have access to a pole to two.
13 This recommendation relates to the limitations stated aove in the su+section on esults(
-
8/13/2019 Bakar Optika Pristupna Mreza
54/101
2.5.2 "ntervention of local authorities as facilitators
#visem &2663(G
K Descries common pulic civil engineering with regard to types of work, current
practice in sharing work &e.g. for Lointly deploying water and electricity( and data
on the volumes of work conducted in France.
K Descries the cases where deployment of ducts is not suitale &wherecommercial Jones have already een connected, where $#T@ networks already
e!ist, where accessile underground galleries e!ist(.
K $larifies requirements to facilitate deployment of a fire+optic network via pulic
civil engineering proLects. "n this conte!t different FTT- architectures are
descried, the selection of numer and siJe of cales is discussed as well as the
need for placing manholes % cainets and the issue of duct dimensioning for a
given numer of operators. Ceneric recommendations are issued for deploying
ducts and dealing with underground manholes.1A
K -ighlights case studies for work done y pulic authorities with different
ackgrounds.
K /pecifies which elements should e taken into account for civil works and whatoptions e!ist for deploying ducts, manholes and duct access. /ample
specifications are provided.
K "dentifies and quantifies the primary cost of deploying ducts together with other
infrastructure. These are ducts, manholes and trench+width+e!tensions &cost
depends on allocation of the trench cost to telecom and the other infrastructure(.
Tale 1G $ost estimates for Loint duct deployment together with other civil
works according to #visem
&1A 8 ducts at *6mm or etter 86mm internal diameter or a single duct of 116mm diameter at minimum.
"deally 8+) ducts of at least )6mm of which 2 are of larger type should e deployed .(
K #ssesses the opportunities and risks of deploying ducts together with other civil
works and descries different strategies with increasing proactive involvement
y pulic authorities.K Descries the approach to implementation through * phases &definition, annual
planning and concrete handling per civil works site(.
2.6 AT earne/ FTT* &or 7reece
#T earney had een commissioned y the -ellenic
-
8/13/2019 Bakar Optika Pristupna Mreza
55/101
K The first stage comprises the provision of fire &the passive network(, i.e. a
FTT- ased infrastructure from the homes to central offices is deployed and
maintained, ending there ehind the Optical Distriution Frame &ODF(. "n the
study this stage is called H"nfrastructure ?roviderI.
K The second stage comprises the provision of wholesale access services, e.g.
wholesale itstream access, y operating O0Ts and O'Ts and aggregatingswitches% routers in the customer premises and in the central offices, respecttively.
This itstream may e as well aggregated at a level aove the central
office. "n the study the second stage is called H$ommunication ?roviderI.
K The third stage is defined y the actual provision of services to end users. This
stage is called H/ervice ?rovidersI.
These elements are comparale to those eing under consideration in /ingapore, see
section *.*.&1; /ee #T earney et al. &266A((.
Dn the 0asis of an analysis of European fi0re deployments9 the study identifiesthreealternative service models
-
8/13/2019 Bakar Optika Pristupna Mreza
56/101
0ased on their e$istin) copper net
-
8/13/2019 Bakar Optika Pristupna Mreza
57/101
In the Wholesale rovider service model !8usiness model ;" and ta/in) thensand
Thessaloni/i to)ether the cumulated unlevered cash flo< turns positive in ;6;2underthe descri0ed assumptions* With the entities separated the Infrastructureroviderservice model !0usiness model :" yields cumulated unlevered cash flo< turnin)positivein ;6;@9
-
8/13/2019 Bakar Optika Pristupna Mreza
58/101
developin) a strate)y to 0oost hi)h speed 0road0and access in rural areas*Fust recently an interministerial committee on pu0lic and private sector proectshasapproved the &TT$ proect* Dn Septem0er 59 ;6639 the %ree/ >inister of
Transport andCommunications has outlined the main dimensions of the proect'+0 (irect fi0re access !&TT$" 0road0and net
-
8/13/2019 Bakar Optika Pristupna Mreza
59/101
instances
-
8/13/2019 Bakar Optika Pristupna Mreza
60/101
re)ards promotin) competition and predicta0ility 0y #Rs as the 0est incentiveforefficient investment in #%*Dn the 0asis of techno-economic scenarios for #%9 the ER% has analysed theimplications and challen)es of #% developments to the Re)ulatory &rame
-
8/13/2019 Bakar Optika Pristupna Mreza
61/101
Relevant remedies or (&s andcontinuation of G up to this point or in certain )eo)raphic areas ar/et ::9 as a ar/et :5" or as a separate 0ac/haulmar/etM (uct sharin) could 0e imposed as an ancillary service to >ar/et ::M 8itstream access to reflect the ne< #% architecture and to allo< for theprovision of hi)h ?uality services*
The tor)an !;662" report analyses the mar/et impacts of incum0entsJ B(S!andAor&TT+" deployment* s the main focus9 the report presents the challen)e forEuropeaninfrastructure-0ased alternative operators* In particular9 the potential responsesofalternative operators are evaluated on the 0asis of a 0usiness model* &or thatpurpose
the approach of replicatin) the incum0entsJ B(S deployment and the option of0ypassin) the incum0entJs net
-
8/13/2019 Bakar Optika Pristupna Mreza
62/101
-
8/13/2019 Bakar Optika Pristupna Mreza
63/101
dependent on the num0er of street ca0inets related to each >(&* In %ermany!factor of76" they could 0e t
-
8/13/2019 Bakar Optika Pristupna Mreza
64/101
infrastructure* Even pay0ac/s of si$ years are possi0le
-
8/13/2019 Bakar Optika Pristupna Mreza
65/101
ducts*>unicipalities play a lar)e role in ri)hts of
-
8/13/2019 Bakar Optika Pristupna Mreza
66/101
constitutional or political reasons9 providin) more autonomy to net
-
8/13/2019 Bakar Optika Pristupna Mreza
67/101
!7" (evelop a frame
-
8/13/2019 Bakar Optika Pristupna Mreza
68/101
;**;*: Technolo)y and net0ps do0psupstream9 provided the copper line len)th 0et !(Sccess >ultiple$er" in the street ca0inets and the end customer is less than 746metres* The ma$imum capacity decreases stron)ly and nonlinear
-
8/13/2019 Bakar Optika Pristupna Mreza
69/101
up)raded 0y installin) fi0re lines 0ets9 0yenlar)in) the street ca0inets space and 0y servin) it ain (istri0ution &rame" locations to 0e phased out if there isnofurther need to maintain a copper line net
-
8/13/2019 Bakar Optika Pristupna Mreza
70/101
desi)ned as an electronic communications medium* The study states it to 0e tooearlyto assess the future role of o
-
8/13/2019 Bakar Optika Pristupna Mreza
71/101
addressed 0y one fi0re strand startin) at the D(& location* This sin)le fi0restrand istransformed into one fi0re per su0scri0er at one !or several cascaded" splitter
!;; The study does not mention &TT8 solutions9
-
8/13/2019 Bakar Optika Pristupna Mreza
72/101
Figure 2;opoloies for :7! fibre net=orks accordin to 7E/
The study points out that 0oth &TT+ ; and D# solutions need fi0re for in-houseca0lin)9 i*e* they cannot use e$istin) !copper" ca0lin)
-
8/13/2019 Bakar Optika Pristupna Mreza
73/101
to providin) multiple on-demand streams !e*)* video on demand"* 8othtechnolo)ies areclassified as not 0ein) future proof concernin) the 0and
-
8/13/2019 Bakar Optika Pristupna Mreza
74/101
K "f there are sufficient fire strands availale, a separate fire to deliver $ale T@
to the end customer might provide a real choice for the customer etween $ale
T@ and "?+T@ offers.
/umming up this comparison there is a preference for fire ased FTT- networks,
while there is a clear preference of usiness customers for the ?2? fire network. From
a regulatory point of view there is a preference for ?2? fire networks as well, ecausethey offer more options for regulatory measures like '' and 9:# where these are
deemed necessary.
2.".2.2 Business models and inestment decisions
The study analyses several usiness models and investment decisions from the
perspective of oth an incument and competitors.
$ale operators
$ale operators can migrate to DO$/"/ *.6 and e!pand the fire closer to the end
customer, thus climing up the ladder of investment, ut they normally have limited free
cash flow. These operators act as a second operator unless they are the first mover or
refuse to invest at all.Telco incuments
"ncuments address the increasing need for more roadand in most cases y rolling
out @D/' networks and, thus, ringing fire closer to the end customer. This limits the
investment compared to an all fire access network. #n FTT- network may e a
second step to enlarge the fire reach when the andwidth demand increase Lustifies it.
This strategy, however, ears the risks that &1( the e!isting large free cash flow may
diminish with falling voice revenues, so that liquidity is reduced when the second step is
necessary and &2( a pure fire network may have a different cost optimal topology thanthereconstruction of the e$istin) copper netoreover9 the incum0entmay
0e 0ypassed 0y a competitor*Some incum0ents already invest in all fi0re access net
-
8/13/2019 Bakar Optika Pristupna Mreza
75/101
loop un0undlin) to move to su0-loop un0undlin) to invest alon)side theincum0entP*
The ne< entrants can neither access the already e$istin) infras