Вы находитесь на странице: 1из 11

Optical Switching and Networking 10 (2013) 89–99

Contents lists available at SciVerse ScienceDirect

Optical Switching and Networking


journal homepage: www.elsevier.com/locate/osn

Protection cost evaluation of WDM-based next generation optical


access networks
Mozhgan Mahloo a,n, Carmen Mas Machuca b, Jiajia Chen a, Lena Wosinska a
a
Royal Institute of Technology KTH, School of Information and Communication Technology, Isafjordsgatan 22, Electrum 229, 164 40 Kista, Sweden
b
Technische Universität München (TUM), Germany

a r t i c l e i n f o abstract

Available online 20 July 2012 New technologies and advanced network devices make it possible to move towards
Keywords: high capacity access networks able to satisfy the growing traffic demand. Wavelength
Next generation optical access networks division multiplexing (WDM) is considered as one of the promising technologies for the
Wavelength division multiplexing next generation access networks since it offers higher bandwidth and longer reach
Passive optical networks compared to the current technologies (such as time division multiplexing (TDM) based
networks). However, the migration to a new technology is typically based on an overall
techno-economic study which should assure the network operator that the new
implementation is cost effective and profitable while able to provide the required
services to the users. Another important aspect in the access network design is the
network reliability performance, which can be improved by providing a certain level of
protection for equipment and/or infrastructure with high failure impact ratio in order to
prevent a big number of the users being affected by a single failure. The cost of
protection should be carefully evaluated since providing the backup resources may be
too expensive for a network operator.
In this paper, we investigate the capital and operational expenditures for two next
generation optical access (NGOA) networks based on the WDM technology in dense urban
areas. Three scenarios with different splitting ratios are studied for each technology, with
and without protection. The aim of this work is to investigate the impact of providing
protection on the total cost of NGOA networks. The results show that in the dense urban
areas the fibers and digging costs are highly shared among the end users but still vary
according to the splitting ratios for different scenarios and the fiber layout. It also can be
seen that with a proper fiber layout design, minor extra investment for protection of
NGOA networks can make a significant saving on failure related operational cost and that
operational expenditures depend significantly on the fiber layout.
& 2012 Elsevier B.V. All rights reserved.

1. Introduction networks to provide a high sustainable bandwidth (BW)


per user. Moreover, applications such as Telemedicine that
The increasing number of internet users and introduc- require an uninterrupted network connection imply an
tion of the new applications such as HDTV, 3DTV and online access network with very high reliability performance. Also
gaming, push service and network providers toward a new the need to reduce the operational cost of the network
design for the next generation optical access (NGOA) moves operators toward node consolidation, which leads to
the reduction of the central offices (CO).
n
Consequently, NGOA networks are expected to cover
Corresponding author. Tel.: 46 87904075; fax: 46 87904090.
large areas (i.e., long reach and high client count), while
E-mail addresses: mahloo@kth.se (M. Mahloo),
cmas@tum.de (C. Mas Machuca), jiajiac@kth.se (J. Chen), providing sustainable bandwidth of 1 Gbps and beyond
wosinska@kth.se (L. Wosinska). for each user as well as an acceptable level of the service

1573-4277/$ - see front matter & 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.osn.2012.07.002
90 M. Mahloo et al. / Optical Switching and Networking 10 (2013) 89–99

reliability. Passive optical networks (PONs) are the most paper we focus on two types of NGOA networks, namely
promising candidates for the future of the access ultra-dense WDM (UDWDM) and hybrid TDM/WDM PON
networks due to its simplicity, transparency and low (HPON), to be deployed in dense urban areas. Introduction
power consumption. PONs can be based on the TDM, of the WDM technology allows for an increase in the
WDM or a combination of both technologies. We envisage number of end premises per PON and total data rates in
that in many cases the NGOA networks will be based on both architectures compared to available optical access
the WDM technology which allows higher capacity per networks such as gigabit-enabled PON (GPON). Adding a
end user. Also due to the large number of the internet TDM layer on top of the WDM offers a better bandwidth
users, new technologies should support large splitting usage in case of HPON. In UDWDM the second splitting
ratios, especially in dense urban areas. step supports a higher number of optical network units
The candidate technology considered for NGOA (ONUs) per PON compared to pure-WDM architectures,
networks should be feasible to implement and operate from while still providing separate wavelengths to each ONU.
the cost perspective. Since the cost sharing in access We investigate three scenarios with different splitting
network is low compared to aggregation and core networks, ratios and number of served buildings per PON to study
cost is an important factor for network providers consider- both capital and operational expenditures (referred to as
ing the limited budget of the residential users. Therefore a CAPEX and OPEX) associated with deployment and opera-
comprehensive cost study is needed before migrating to the tion of such networks. Here we provide a comprehensive
new technologies. Furthermore, future access network relia- cost study for the whole access network from head end
bility performance will be an important issue, because they network in the central office to the end user promises,
will cover large service areas as a result of node consolida- considering the network life time, network planning and
tion and long reach [1]. Hence the need for the failure fiber layout. Moreover the added cost of the protection till
management and protection cannot be ignored. Although first splitting point is investigated and compared with the
the protection down to every end user is not targeted yet by case without any protection.
network operators due to the high costs, the protection of The remainder of this paper is organized as follows.
some components and/or infrastructure is foreseen so that Section 2 describes two variants of NGOA architectures.
the failures with higher user impact are reduced. Section 3 introduces the scenarios and the fiber layout
Recently a large number of studies have been done considered in this study. Section 4 presents the cost
aiming to propose a good candidate for the next generation classification and the values used for the CAPEX and OPEX
of the optical access networks with emphasis on WDM calculations. In Section 5, we evaluate the results of the
based technologies [1,3–6]. The main focus is to find an cost study for different variants of the NGOA in dense
option supporting high sustainable bandwidth and long urban areas. Finally, Section 6 gives some conclusions.
reach, as well as allowing for the increasing number of the
internet users [2,3]. However, different requirements of the
operators and end users as well as the lack of a single 2. Next generation optical access networks
mature technology, prevent the networking society to agree
on a particular technology for the future access networks. A The considered NGOA networks are based on the well
comprehensive cost study of the proposed technologies known PON architecture (see Fig. 1), which is a passive
considering the operational and capital expenditures in a point-to-multipoint network where the head-end device
complete scenario together with the infrastructure cost is located at the central office (CO) is called the optical line
an interesting research area that can provide a better view terminal (OLT), and the device at the user side is referred to
to network operators and facilitate their migration path as the optical network unit (ONU). The remaining part of the
toward suitable NGOA architecture. Therefore, some efforts network is the fiber infrastructure and the passive compo-
have been put on the cost analysis and feasibility of the nents at the remote node (RN). The fiber infrastructure
new technologies, providing the techno economic study of consists of cables with several optical fibers. The optical
the NGOA candidates [7–11]. However, these articles are fiber interconnecting the OLT with the RN is referred to as
typically involved with the assessment of initial investment the Feeder Fiber (FF), whereas the optical fiber interconnect-
needed to set up the network and/or power consumption ing the RN with the user is called Distribution Fiber (DF).
[12,13] of access networks candidates while cost for oper- Most of the currently deployed PONs are TDM based having
ating the network is only addressed in the scenarios with a power splitter at the RN. A few WDM PONs with a
relatively low bitrate per customer and short reach. In this wavelength demultiplexer (e.g., an array waveguide grating
paper we target truly future proof NGOA networks offering (AWG)) at the RN are also deployed. PONs are limited by the
high bit rate and covering large service areas. Moreover, the
related papers are normally comparing performance of
different technologies, while in this work we are also
examining the effect of different splitting ratios, in the total
cost of ownership for each NGOA candidate as well as
assessing the extra cost of adding protection in each of the
options.
In the cases where a very high dedicated bandwidth
and client count per fiber are desired, there are only a few
NGOA candidates to support these requirements. In this Fig. 1. Traditional passive optical network.
M. Mahloo et al. / Optical Switching and Networking 10 (2013) 89–99 91

Fig. 2. Considered WDM-based NGOA architectures: (a) HPON and (b) UDWDM.

client count per feeder fiber and/or the passive reach Table 1
(distance from CO to the end user). These limitations are Considered architecture options.
overcome by NGOA solutions that can support around 1000
Architecture Number of Number of Number of
clients per FF and up to 100 km theoretical reach. variant AWG output splitter output ONUs per FF
In this study, two NGOA architectures have been ports in RN1 ports in RN2
evaluated: based on the hybrid TDM/WDM passive optical
network (HPON), and based on the ultra-dense WDM HPON 1:40/1:8 40 8 320
HPON 1:40/1:16 40 16 640
passive optical network (UDWDM PON) (see Fig. 2). HPON 1:80/1:16 80 16 1280
UDWDM 1:40/1:8 40 8 320
2.1. Hybrid TDM/WDM passive optical network (HPON) UDWDM 1:80/1:8 80 8 640
UDWDM 1:20/1:32 20 32 640
The considered HPON (shown in Fig. 2(a)) combines
TDM and WDM schemes. The advantage of having WDM is
the increase of the data rate offered to each user; while the
advantage of TDM is its flexibility in bandwidth utilization. 2.2. Ultra-dense WDM passive optical network (UDWDM)
This architecture has a tree topology with the OLT on the
top of the tree. The OLT is connected to the first remote This architecture is able to offer at least 1 Gbps bit rate
node (RN1) that contains a passive wavelength splitter/ per channel over tens of kilometers and thus, it can meet
combiner (e.g., AWG) and its outputs go to the second the requirements of next generation fiber to the building
remote nodes (RN2s). The AWG separates channels so that (FTTB) networks studied in this paper [1]. Each ONU is
only one single wavelength will reach each RN2, so there are assigned its own wavelength channel and is physically
as many RN2s as wavelengths in the WDM signal. Each RN2 separated from other users by means of a virtual point-
is a power splitter with every output going to one of the to-point connection. The architecture has a tree topology
ONUs. The downstream traffic carried by each wavelength is with the OLT on the top of the tree which is connected to a
broadcasted to all the users connected to the same RN2. In first remote node (RN1) that contains a passive waveband
this architecture, the ONUs do not have dedicated wave- splitter/combiner (demultiplexer/multiplexer) based on
lengths since all ONUs connected to the RN2 use the same arrayed waveguide grating (AWG). Each RN1 output port
wavelength and apply TDM technique for the upstream goes to a second remote node (RN2) which includes a
traffic. In order to support high sustainable bandwidth per passive power splitter/combiner. Each RN2 is a power
user, HPON requires 10 G transceivers in both the OLT and splitter with every output going to one of the ONUs. The
the ONU side. Using a 10 G transceiver as well as the need of band splitter divides the whole spectrum received in
managing the TDM makes the ONU complex and expensive its input port into separate wavebands each of which
in this scheme. For this study we compare three different goes to one distinct RN2. One waveband includes several
splitting configurations which are listed in Table 1. wavelengths each of which is assigned to a different ONU.
92 M. Mahloo et al. / Optical Switching and Networking 10 (2013) 89–99

The use of power splitters implies a broadcast and select the Manhattan model [14] so that buildings are grouped
behavior, that is, the signal going through each output of in blocks, which are arranged in an array manner. The
RN1 (i.e. each input of RN2) is broadcasted to all ONUs Manhattan model is a proper geometric model to be
connected to RN2, so that each ONU has to ultimately select considered in dense urban areas since it is well mapped
its assigned wavelength from the received waveband. into some highly populated areas in big European cities
Furthermore, we take into account wavelength chan- such as the Eixample district in Barcelona.
nel capacity of 1 Gbps in order to support the required We assume blocks with 8 buildings on each side, i.e. 32
data rate per user. For this study, three different splitting buildings per block. Based on the technological reach
ratios have been considered, each of them differing on limitation for each architecture [3], the maximum num-
maximum reach and therefore, on number of covered ber of blocks that can be covered by one central office can
users by a single CO (see Table 1). be computed using formula (1) where maxReach is the
maximum reach for a given technology and splitting
2.3. Protection scheme ratios [15], Blocksize (see Table 3) is the length of one
block in kilometer, Vblock and Hblock denote the number of
The feeder fiber and the RN1 are the parts of the vertical and horizontal blocks respectively. As an example,
network with the highest impact of a single failure affecting if we consider a service area depicted in Fig. 4, the longest
a very big number of supported connections (from 320 to dashed line in the outer border of the area should be
1280 ONUs in our studies). The failure rate of RN1 is much equal or less than the maximum reach, with Hblock equal
lower than FF and therefore, the considered protection is to 5 and Vblock/2 equal to 3.
based on adding a second disjoint FF. In the protection
scheme, 2:N AWG replaces 1:N one to support both work- max Reach V
4ðHblock þ block Þ ð1Þ
ing and protection FFs (as depicted in Fig. 3). Furthermore, Blocksize 2
an optical switch is also required at the OLT side to connect
the line card to the FFs. The rest of the network is identical
for protected and unprotected schemes. Table 2
Architectures main characteristic [15].
3. Network scenarios and proposed fiber layouts
Architecture option Max reach Area Number of
[km] [Hblock  Vblock] ONUs
In this study we evaluate the cost and energy con-
sumption of the aforementioned NGOA architectures in a HPON 1:40/1:8 38 200  170 1088.000
dense urban scenario. The maximum reach of each variant HPON 1:40/1:16 16 100  70 224.000
is presented in Table 2. It can be observed that the HPON 1:80/1:16 16 100  70 224.000
UDWDM 1:40/1:8 59 200  300 1920.000
maximum reach is limited by the splitting ratio of the
UDWDM 1:80/1:8 48 200  260 1664.000
second remote node since it is broadcasting the signal to UDWDM 1:20/1:32 27 200  100 640.000
all the output ports. The considered scenario is based on

Fig. 3. Considered NGOA network architectures without and with protection.


M. Mahloo et al. / Optical Switching and Networking 10 (2013) 89–99 93

Then number of ONUs is calculated by multiplying the layout has been proposed that minimizes the FF length.
total number of blocks in the area by 32, which is equal to The working FF (i.e. FFw) interconnecting the OLT and RN1
the number of buildings covered by this CO. The max- through the shortest path is depicted as a solid line. The
imum covered area in terms of blocks (Vblock  Hblock) and FF protection (i.e., FFp) uses a disjoint duct (depicted as
number of ONUs are listed in Table 2 for each particular tick dashed line) which just requires connecting existing
architecture. dark fiber (or pumping fiber when no dark fiber is
In the fiber to the building (FTTB) scenarios (with one available) while avoiding any extra trenching. The dis-
ONU per building), optical fiber should reach every single tribution fiber (DF) which connects RN1 to RN2 is
building and therefore, all streets are assumed to have depicted as a dashed dotted line. The last fiber section
one duct containing all required fibers. The fiber path so-called building fiber (BF) interconnects RN2 with each
between CO and each building is divided to three sections, building and is depicted as an arrow line. All three fiber
as shown in Fig. 4. This figure presents as an example, the sections are sharing the trenches that cover all the streets
proposed fiber layout for the HPON and UDWDM 1:40/1:8 in the area. Each circle represents 8 collocated splitters;
variants. Since network operators typically want to offer where each splitter covers one block side in its corner.
protection down to RN1, in our study a protected FF Fig. 5 shows the model for RN1 distribution and FF
layout of all six scenarios used for the calculation of the
number of fibers passing through each street that affects
the OPEX and also the total fiber and trenching length
Table 3 needed to cover the whole service area which is part of
Considered parameters. the CAPEX calculation.
The circles represent one AWG placed in RN1 connected
Parameter Value Unit
to the CO using the working and protection FF (depicted by
Tech. salary 255 $/h solid and dashed line respectively). As explained earlier,
Block size 0.113 Km each RN1 is connecting to a number of splitters located in
Street size 0.02 Km the street crossings. The number of splitters depends on the
Speed 20 km/h
Energy cost 0.1 $/kWh
splitting ratios. The reason of having different fiber layout
Yearly building rental fee 290 $/m2 for various architectures is the number of ONUs per FF that
defines how many blocks can be served by one RN1. As the
Component Cost Power [W] FIT
AWG 1:80 cost 2250$ 0 200 number of ONUs connected to same AWG increases, the
AWG 1:40 cost 1500$ 0 200 density of the circles (RN1) is decreasing for the same size
AWG 1:20 cost 1125$ 0 200 of the area. For example Fig. 5(a) shows the HPON and
Optical switch 1:2 1000$ 0 250 UDWDM 1:40/1:8 scenarios serving 320 buildings per PON
Splitter 1:N 2.5$/port 0 170
Shelf with 20 slots 4860$ 90 0
that is equivalent to 10 blocks. The middle figure depicts
Fiber 160 $/km 0 570/km the UDWDM 1:20/1:32, 1:80/1:8 and HPON 1:40/1:16 FF
Trenching 60000 $/km 0 0 layout. Finally Fig. 5(c) shows the HPON 1:80/1:16 variant.
HPON Line terminal (OLT) 87.5$/building 1.3–2.9 256
UDWDM Line terminal (OLT) 106 $/building 2.3–2.4 256
HPON ONU 175$ 5.5 256 4. Considered cost classification
UDWDM ONU 145$ 5.6 256

For a given dense urban scenario we compare the cost


of the proposed optical access technologies with and
without protection. The costs considered in this compar-
ison are either capital or operational expenditures.

4.1. Capital expenditures (CAPEX)

 Network equipment, which includes the cost of purchas-


ing the required network equipment, based on the
shopping list resulting from the network dimensioning.
 Equipment installation. The installation cost of the com-
ponents forming the access part of the network including
OLT, ONU and devices located in the RNs is considered
for the calculations, which depends on the number of
technicians required for the installation as well as on the
traveling time to the equipment location and the time
required for the installation. This cost varies per country
since the technician salary may differ a lot according to
the area (e.g., much higher in European countries than
most of the Asian countries such as China).
 Network infrastructure. The network dimensioning
Fig. 4. Network model and proposed fiber layout for 1:40/1:8 solutions. and the fiber layout approach shown in the previous
94 M. Mahloo et al. / Optical Switching and Networking 10 (2013) 89–99

Fig. 5. RN1 distribution and feeder fiber layout. (a) HPON and UDWDM 1:40/1:8, (b) UDWDM 1:20/1:32, UDWDM 1:80/1:8 and HPON 1:40/1:16 and (c)
HPON 1:80/1:16.

section result in a number of fibers and cables that


should be installed at each street. Based on that, the
cost of purchasing the cable and the fibers, as well as
the installation cost can be computed. The installation
cost takes into account the number of splices that
should be done, depending on the cable size. Cost of
the trenching to cover the whole area is also a
considerable part of the infrastructure cost considered
in this part of expenses.

4.2. Operational expenditures (OPEX)

 Failure reparation (FR). The FR cost depends on the


number of expected failures in the network during its Fig. 6. Rack floor space model [16].
operational life. This value depends on the failures in
time (FIT) rate of the installed components. Moreover,
the failure reparation cost also depends on the number presented in Table 2, with the total floor space needed
of technicians required for the reparation, and the MTTR for each rack. The total space for each rack is equal to
(mean time to repair) for each failure. For simplicity, we the sum of the rack’s bottom area (0.6 m  0.3 m) and
have assumed that the MTTR is the mean value of the enough working space for the technicians in front of it
sum of the reparation time of the component and twice (0.6 m  1.0 m) shown in Fig. 6. Each rack can place up
the traveling time to the location of this component, to four shelves with 20 slots in each [16]. This study
neglecting the failure diagnosis time. compares the CO floor space per ONU which is calcu-
 Penalties (P) associated with each connection interrup- lated by dividing the total required CO area for locating
tion impact also the cost of a network. This penalty is the OLT racks, by the number of buildings covered by
usually agreed at the service level agreement (SLA) that CO. Having the yearly rental of the central office
between user and service provider. Up to now, this we can calculate the cost associated to the CO floor
value is included in SLA only with business users, but space and add it to the yearly OPEX per ONU.
in the future it is expected to be extended to the
residential users. SLA includes the connection avail- 5. Cost comparison
ability time which is related to the allowed connection
interruption time during a defined time period. This The cost comparison comprises capital and operational
cost depends on the penalty rate PR ($/h), the percen- expenditures as presented in the following subsections.
tage of business users in the area (in dense urban areas The final comparison combines both aspects for a given
is much higher than in rural areas), and on the number network operational time. The cost evaluation considers
of affected users per a given failure. In this study, two the maximum covered area for each particular architec-
different PRs have been considered: 10 and 100$/h. ture as given in Table 2 and the input data provided in
Since this scenario is a dense urban area, the percen- Table 3. The values for power consumption and FIT are
tage of business users is assumed to be 80% and the directly extracted from [15], while the input for cost are
average traveling speed is 20 km/h. translated in US dollar unit according to the available cost
 Power consumption. The power consumption asso- figures in [15] which are normalized to the common
ciated with each active component of the network baseline of the GPON ONU.
can be summed up and multiplied by the power cost in
every year. 5.1. Capital expenditures (CAPEX)
 Floor space. Several racks are located in the central
office to place the OLT’s equipment such as line cards, The capital expenditures (CAPEX) per building for
power supplies, aggregation switches and so on. We both protected and unprotected scenarios is presented
calculate the total CO area by multiplying the number in Fig. 7. As described in the previous section, the CAPEX
of the racks needed to serve the number of ONUs includes three expenditures: equipment, installation and
M. Mahloo et al. / Optical Switching and Networking 10 (2013) 89–99 95

infrastructure cost. The expenses related to the infra- as well as the expected connection interruption time per
structure has the highest contribution to CAPEX due to ONU have been compared in Fig. 8(a) and (b) respectively.
the fact that trenching cost is so expensive. Furthermore, It can be observed that the failures with highest impact on
the equipment is considerably costly due to the high ONU the reparation time are related to component breakdowns
price that is not shared, while the rest of expenses are (especially ONU failures, which are the most frequent ones)
divided by the number of buildings connected to each rather than fiber cuts. Differences between unprotected and
particular component, which depends on the splitting ratio. protected scenarios regarding the failure reparation (FR)
Generally the component cost is slightly higher in HPON cost are negligible, which is not the case with regard to
compared to the UDWDM caused by the higher price of the connection interruption times. For this parameter a
ONU to support TDM capabilities and 10 G transceivers. The significant decrease of connection interruption times for
small increase of the equipment cost in the protected case the protection scenarios from tens of hours per year to a
is introduced by adding the price of the optical switches at minute per year can be observed (see Fig. 8(b)). In this graph
OLT side. Installation cost numbers differs the most among it can also be shown that the reason of the high interruption
different variants, which is mostly related to the splitting time for unprotected architectures is the number of fiber
ratio of the second remote node (RN2). The higher splitting failures due to the fact that a single cable cut affects many
ratio in RN2 the less number of splitters needed and users (compared with the ONU failure which affects only
consequently higher cost sharing among buildings. one single building).
The infrastructure cost is highly dependent on the When aiming at comparing operational costs, Fig. 9
splitting ratio and the fiber layout. Feeder fiber expenses shows for each architecture and protection scenario the
are the differentiating factor between the various scenarios. failure reparation (FR), penalty, power and floor space
In both architectures the highest fiber related cost corre- costs per ONU per year. It has to be mentioned that the
sponds to the 1:40/1:8 case due to the larger supported penalty rates considered in this study are: 10 and 100 $/h.
area. It also can reflect the influence of AWG placement in It can be seen in Fig. 9 that the highest operational
the area. Generally the CAPEX numbers are dependent on costs of unprotected networks are the penalty costs due to
different factors such as number of ONUs per FF, and the failures in the infrastructure. The protection of the feeder
splitting ratio of RN2 which implies higher possible sharing fiber reduces significantly the penalties (around three
of the expenses among buildings. It is clear from the figure orders of magnitude) while almost no changes have been
that the lowest number is belonging to the UDWDM 1:20/ observed in the other costs. The impact of the FF layout on
1:32 supporting 640 users per PON. The added cost of the the penalty costs can be also observed. Looking at Fig. 5(a)
protection up to the first level is considerably low in all the it can be seen that the FF layout is rather distributed
cases, though the numbers are slightly higher in 1:40/1:8 through all the streets and a fiber cut will surely affect
variant of both NGOA architectures, again caused by less FF(s) while in Fig. 5(c) only a few streets contain FF cables
number of ONUs served by one feeder fiber. and therefore only few cable cuts will involve FF failures.
This difference makes that the penalties of networks with
5.2. Operational expenditures (OPEX) a distributed FF layout are higher than of architectures
using a more concentrated FF layout (shown in Fig. 5(c)).
As a first step towards the computation of the opera- The failure reparation costs are slightly higher due to the
tional expenditures, the failure reparation required time increase of amount of fibers and components (i.e., optical

Fig. 7. CAPEX per ONU distribution for protected and unprotected scenarios ($/ONU).
96 M. Mahloo et al. / Optical Switching and Networking 10 (2013) 89–99

Fig. 8. Failure reparation time (a) and connection interruption time per ONU (b) per year [hour/year].

switches) required for protection. Furthermore, the cost of at these streets leading to much less FFs failures and
electrical power consumed by the active equipment is a hence, lower penalties and less failure reparation than in
considerable part of the OPEX, while there is no power HPON 1:40/1:8 (the cables containing FF have higher
usage related to the infrastructure. Compared to the other number of fibers than the other cables).
aspects, floor space has relatively lower contribution to Fig. 10 also shows that the impact that infrastructure
the OPEX especially for HPON variants. It should be noted failure reparation has on the total cost is rather low. The
that the OLT for HPON needs less space compared to the reason is that in a dense urban area a large number of
UDWDM PON. The reason is that in UDWDM each build- users are covered by few tens of optical cables and hence,
ing is served with a separate wavelength which means the failure reparation is highly shared. Furthermore,
one transceiver is needed for each ONU (normally trans- penalty costs of protected networks are significantly
ceivers array is used instead of individual ones), while in lower than the unprotected counterparts, which justify
the case of HPON each wavelength is shared by a number the small investment required to provide protection.
of ONUs depending on the RN2 splitting ratio. Therefore Fig. 11 shows the total cost per ONU for the different
the office size needed for the UDWDM is at least twice as architectures for 20 years. It can be observed that
big as the one for HPON, which makes the hybrid PON UDWDM 1:20/1:32 version has the lowest cost except
more cost effective in the node consolidation scenarios for the unprotected case where operator is charged by
where the number of the active central offices is reduced. business users for the service interruption penalty. This is
due to the lower CAPEX and energy consumed per ONU
5.3. Total cost of ownership (TCO) compared with other architectures. On the other hand,
the unprotected UDWDM 1:20/1:32 has moderate FF
The total cost per building for a network operational failure probability(FF are laid in some of the vertical
time of 20 years is shown in Fig. 10. It can be observed streets) as shown in Fig. 5(b) and therefore, when a
that for unprotected solutions, penalty costs are compar- penalty rate is applicable, the cost increases significantly
able with the infrastructure costs and in some cases (e.g., more than if it had other FF layout with lower failure
UDWDM or HPON 1:40/1:8) even higher than the total reach. This is the case of HPON 1:80/1:16, which has the
CAPEX. It should also be mentioned that the penalty cost lowest amount of FF cables laid in vertical streets and
differences between various splitting ratios of the same therefore, it is shown that in unprotected scenarios with
technology are due to the fiber layout. The splitting ratio penalty rates applied, it is the least costly option due to
determines the ONU and the RN clustering and therefore, the low number of unprotected FF cable cuts, as explained
the number of affected connections per FF and DF varies previously.
with the clustering. The more FFs are grouped, the less FF
failures will occur but with higher impact. This effect can 6. Conclusion
be seen for example with HPON 1:40/1:8 and HPON 1:80/
1:16. The former has RNs at each vertical street, whereas This paper has presented a detailed cost comparison of
in the latter RNs are located at few vertical streets. some most promising next generation optical access
Consequently, in HPON 1:80/1:16 FFs are installed only network options with different fiber layouts, considering
M. Mahloo et al. / Optical Switching and Networking 10 (2013) 89–99 97

Fig. 9. Operational expenditures [$/ONU/year] for unprotected (UP) and FF protected (P) scenarios.

Fig. 10. TCO [$/ONU] for a network operational time of 20 years in logarithmic scale.
98 M. Mahloo et al. / Optical Switching and Networking 10 (2013) 89–99

Fig. 11. TCO per ONU for the different architectures and different penalty rates during 20 years [$/ONU].

both unprotected and protected feeder fiber. The compar- Programme (FP7/2007-2013) under grant agreement
ison has been performed not only in terms of CAPEX but no. 249025 (OASE).
also OPEX including failure reparation costs, penalties due
to service disruption, floor space and energy consumption. References
It can be concluded that in dense urban areas the CAPEX is
not differing significantly for various NGOA technologies [1] 249025-ICT OASE Project Deliverable 2.1, Requirements for
and their variants, due to the high sharing cost among European Next-Generation Optical Access Networks.
[2] R. Hülsermann, C. Lange, Topology analysis of the access/aggregation
users. Generally the 1:40/1:8 configuration of both tech-
network structure in future optical access networks, in: Proceedings
nologies has the highest TCO (CAPEX and OPEX), caused by of the 11th ITG Conference on Photonic Networks, Leipzig, Germany,
the less number of client counts per FF. The presented May 03–04, 2010.
results show the importance of offering protection of the [3] 249025-ICT OASE Project Milestone 4.2, Assessment of Next
Generation Optical Access System Concepts.
feeder fiber in order to significantly decrease the TCO per [4] M. De Andrade, G. Kramer, L. Wosinska, J. Chen, S. Sallent,
ONU when penalties are applied. Another result of this B. Mukherjee, Evaluating strategies for evolution of passive optical
paper is the influence that the fiber layout (in particular the networks, IEEE Communications Magazine 49 (7) (2011) 176–184.
FF layout due to its high failure reach) has on the penalty (Juliol).
[5] J. Prat, J.A. Lázaro, K. Kanonakis, I. Tomkos, New FTTH architectures
cost: it can be observed that the less streets containing FF for NG-PON-2, OSA Access Networks and In-house Communications
(i.e., the more concentrated FF layout), the lower the (ANIC) Topical Meeting, June 21–24 2010, Karlsruhe, Germany.
penalty costs. Moreover it has been shown that for the [6] A.M.J. Koonen, N.C. Tran, E. Tangdiongga, The merits of reconfigur-
ability in WDM-TDM optical in-building networks, in: Proceedings
proposed fiber layout the increase of investment for offer-
of the Optical Fiber Communication Conference (OFC) 2011, 2011,
ing protection is a low percentage of the TCO. When no pp. JWA63.
penalties are considered, the TCO increases less than 2.5% [7] D. Breuer, C. Lange, E. Weis, M. Eiselt, M. Roppelt, K. Grobe,
when offering FF protection while when penalties are J.P. Elbers, S. Dahlfort, F. Cavaliere, D. Hood, Requirements and
solutions for next-generation access, ITG Conference on Photonic
applied, up to 95% of the TCO can be saved. It may justify Networks, May 2011, Leipzig, Germany.
the investment for installation of disjoint FF when penalties [8] N.C. Tran, H.D. Jung, C. Okonkwo, E. Tangdiongga, T. Koonen, A
are applicable. In this work we used Manhattan model as 10 Gb/s Passive-Components-based WDM-TDM Reconfigurable,
Optical Access Network Architecture, in: Proceedings of the Optical
starting point to show a general picture about the differ-
Fiber Communication Conference (OFC) 2011, 2011, pp. OThT1.
ences of the architectures. This study will be extended by [9] C. Mas Machuca, J. Chen, L. Wosinska, Impact of protection to
using our geographical network modeling tool to have capital and operational expenditures of optical access networks,
more realistic cost comparison in real scenarios. Informations Technische Gesellschaft im VDE (ITG), May 2011,
Leipzig, Germany.
[10] J.K. Grobe, M. Roppelt, A. Autenrieth, J.P. Elbers, M. Eiselt, Cost and
energy consumption analysis of advanced WDM-PONs, Commu-
Acknowledgment nications Magazine, IEEE 49 (2011) s25–s32.
[11] C. Mas Machuca, M. Mahloo, J. Chen, L. Wosinska, Protection cost
evaluation of two WDM-based next generation optical access
The research leading to these results has received fund- networks, in: Asia Communications and Photonics Conferen-
ing from the European Community’s Seventh Framework ce(ACP), November 2011, Shanghai, China.
M. Mahloo et al. / Optical Switching and Networking 10 (2013) 89–99 99

[12] K. Grobe, M. Roppelt, M. Eiselt, J.P. Elbers, Combined reach, client- [14] M.A. Marsan, G. Francese, F. Neri, Manhattan topologies for passive
count, power-consumption, and cost analysis of WDM-based all-optical networks, in: 9th Annual European Fibre Optic Commu-
next-generation PON, ECOC 2011, September 2011, Geneva, nications and Local Area Network Exposition, June 1991, London, UK.
Switzerland. [15] 249025-ICT OASE Project Deliverable 4.2.1, Technical Assessment and
[13] Christoforos Kachris, Ioannis Tomkos, Power Consumption Evalua- Comparison of Next-Generation Optical Access System Concepts.
tion of Hybrid WDM, PON Networks for Data Centers, Athens [16] OASE Project Deliverable 5.2, Process Modeling and First Version of
Information Technology, Athens, Greece. TCO Evaluation Tool.

Вам также может понравиться