Professional Documents
Culture Documents
TOC of 5G Fronthauling
TOC of 5G Fronthauling
TOC of 5G Fronthauling
1. INTRODUCTION
Mobile data traffic envisioned for the fifth generation (5G) of cellular networks sets extremely challenging
requirements for the transport networks, which are used for aggregating traffic from the base-stations (BSs) to
the mobile packet core. These requirements are specified in terms of high capacity and throughput, increased
reach, reconfigurability, reduced latency, and power consumption. To meet these requirements at a reasonable
cost, several issues will arise, especially regarding a carbon footprint due to high power consumption and mobile
traffic tsunami [1], i.e., a seven-fold traffic growth has been forecasted for the 2016 – 2021 period with
a compound annual growth rate of 47% [2]. Moreover, this traffic increase will intensify because of numerous
emerging services emerging such as Internet of Things (IoT), and gigabit wireless connectivity, just to name
a few. As a result, 5G cellular network should be able to accommodate capacities as high as 20-Gbps per cell,
throughput values as high as 1-Gbps per user, and latency as low as 1-ms [3]. Such extremely ambitious goals
are requiring significant technological advances in both wireless and transport network infrastructures. These
advances must be provided while making sure they have a limited impact on both cost and carbon footprint [4].
Despite this, most of the works about optical fronthaul focus on capacity [5], [6] and latency [7], [8] issues,
while the cost implications of the proposed fronthaul solutions remain unexplored. Up to now, very little
considerations have been devoted to the total cost of ownership (TCO) of fronthaul networks that are based on
different radio-over-fiber (RoF) architectures [9]. Such knowledge could lead to more informed decisions when
network deployment or migration strategies need to be developed.
Therefore, in this paper, first, we provide a short overview of the RoF architectural options that could be used
in a fronthaul network, and, second, we describe a TCO model together with the assumptions used for the cost
assessment. Finally, we provide a comparison between capital and operational expenditures (CAPEX, OPEX)
for the considered fronthaul solutions, employing intermediate frequency (IF), digital signal processing assisted
(DSP-A), and digital (D) RoF techniques for interconnecting cell-site and baseband equipment when a greenfield
scenario considering both fiber deployment and fiber leasing is examined.
an RRU and a BBU uses the common public radio interface (CPRI) [10], [11] or the open BS architecture
initiative (OBSAI) [12] the fronthaul data rate will be well-beyond 1-Tbps [4], [6].
BBU pool
Core-metro
network
BH MFH
Carrier (de-)
ME Carrier (de-) aggregation
aggregation
Figure 1. Block diagram of mobile traffic backhauling based on fiber-optic links and supporting carrier
aggregation. Acronyms: ME – metro edge, BH – backhaul, BBU – Baseband unit, MFH – mobile fronthaul,
RRU – Remote radio unit, and MIMO – multiple input multiple output.
Theoretically, meeting such high bandwidth requirements with existing fronthaul technologies requires the
installation of new optical cables or very expensive high-speed transceivers [4], [13]. Ultimately, service
providers will spend more and more money for their mobile access network infrastructures. This is not
a sustainable solution, especially if cost-effectiveness and carbon footprint are the concerns [13], [14]. Therefore,
alternatives for RRU-BBU interfacing based on analog radio-over-fiber (A-RoF) have been considered. Note
that the functional splits provided by the eCPRI specification are not considered in this paper for a sake of
clarity. These aspects are left for the future work.
In A-RoF solutions, the aggregated wireless signals, which modulate an optical carrier, may have their spectral
bandwidths unchanged, leading to a relatively high bandwidth efficiency as compared to D-RoF. In fact, the
transmitted optical signal is an analog signal with a spectral efficiency much higher than for intensity modulated
signals used in CPRI. In addition, the bandwidth efficiency of an A-RoF solution may further be enhanced via
(1) frequency-division multiplexing (FDM) through analogue signal processing that includes frequency
conversion, and RF combining/splitting [9], [14], or (2) wireless channel (de-) aggregation involving DSP
techniques that are based on fast Fourier transformation (FFT) and inverse-FFT (IFFT) [8]. The first solution is
referred to as intermediate frequency over fiber (IF-A-RoF), while the latter to as DSP-assisted A-RoF (or DSP-
A-RoF). These options are promoted as efficient mobile fronthauling techniques since they provide solutions
with high flexibility, capacity, and scalability while supporting 5G requirements and reducing capital and
operational expenditures. However, both have serious drawbacks. IF-A-RoF has a high susceptibility to
transmission impairments [6], [14]. DSP-A-RoF has latency issues [7], [8] since the mobile channels are
aggregated and then de-aggregated in the digital domain using digital signal handling techniques based on
FFT/IFFT operations.
Infrastructure
management
fiber leasing
Energy cost
Floor space
Backhaul Backhaul
Purchasing
Installation
Faul
deployment migration
Total cost of
ownership
Figure 2. Assessment of a total cost of ownership: (a) model and (b) flowchart.
2
ICTON 2018 We.C2.5
It consists of two stages – wireless deployment and transport network deployment or migration. Since this
work focuses on the TCO assessment of the mobile fronthaul (MFH) network, costs related to the wireless
deployment and spectrum leasing are not included in the analysis. They remain the same for all three RoF
options considered for the BBU-RRU interface. A fronthauling can be realized using one of two alternatives –
greenfield deployment and brownfield deployment or migration. In the context of optical aggregation networks,
a greenfield deployment means that the fiber infrastructure is absent and must be deployed from scratch. In turn,
a brownfield scenario assumes that fiber ducts are already installed, and new fibers may be pumped into a duct
when needed. In a brownfield scenario the trenching cost is not accounted for and, as a result, increasing the
link’s capacity it is easier and less costly. The brownfield option is not considered in this paper, and it is left to
be considered for a future work.
(a) AE AE (b)
AE AE
AE
ME
AE
AE AE
AE AE AE AE
MN MN
AE
AE AE AE AE
AE AE
AE
MN
AE
AE AE
AE AE AE AE
AE AE
AE
AE
AE AE AE AE
Figure 2. Scenario and assumptions considered for the TCO assessment: (a) mobile fronthaul (MFH) network
topology and (b) the list of assumptions. Acronyms: ME – metro edge, MN – Metro node, and AE – access
edge.
For the TCO assessment, we consider two scenarios – greenfield fiber deployment and greenfield fiber leasing.
The latter means that we include another term in the OPEX calculation taking into account the leasing price per
fiber and per kilometer. The network scenario under exam assumes a dense-urban network, typical of large cities
such as Tokyo or New York. The configuration of the MFH network is illustrated in Fig. 3 (a). It comprises one
metro ring, i.e., connecting three metro nodes (MN), one metro edge (ME), and seven access rings (AR) with
five access edges (AE) each. Each AE is co-located with a BS with 200 three-sector antennas in a tower that is
also connected to its corresponding RRU. BBUs are located at the corresponding MN. In case of high BBU load,
we assume that baseband processing can be performed at a neighboring MN or at the ME. Therefore, each MN is
equipped with passive optical switches based on arrayed waveguide gratings (AWGs) to redirect a lightpath, and
with optical amplifiers (e.g., Erbium doped fiber amplifiers – EDFA) and to compensate for possible optical
losses. The other numerical assumptions required for the TCO analysis are given in Fig. 3 (b), while a more
detailed explanation is provided in [4]. In addition, the costs of the most significant components/equipment
considered during the assessment are summarized in Table 1 at the end of the paper.
3
ICTON 2018 We.C2.5
on the overall cost of the CPRI-based architecture. Surprisingly, “fiber deployment” has a minor impact on
CAPEX in a greenfield deployment. On the contrary, if fiber leasing is considered, then the leasing cost has
a significant impact on the OPEX value. For the fiber leasing case, maintenance costs are almost half of the
OPEX regardless of the number of carriers that are aggregated. For the greenfield deployment, maintenance
costs are dominant. Finally, if one compares how CAPEX and OPEX change with number of aggregated carriers
for the different MFH architectures, it becomes obvious that, at a certain number of aggregated carrier, the
CAPEX/OPEX values of DSP-A-RoF become equal to the CAPEX/OPEX values of IF-A-RoF and/or D-RoF,
i.e., there are some “crossing points” on the plot showing how cost changes with the number of aggregated
carriers.
D-RoF
D-RoF
4 32 4 32 4 32 4 32 4 32 4 32
SCM SCM
(c) ×104 (d) ×104
DSP-A-RoF
IF-A-RoF
IF-A-RoF DSP-A-RoF
D-RoF
D-RoF
4 32 4 32 4 32 4 32 4 32 4 32
SCM SCM
Figure 3. Capital and operational expenditures with cost differentiation according to the category, assuming
a greenfield fiber deployment and fiber leasing for the IF-A-RoF, DSP-A-RoF and D-RoF fronthaul
architectures when 4 and 32 carriers are (de-)aggregated at a remote radio unit (RRU), CU = 2500 USD.
5. CONCLUDING REMARKS
A total cost of ownership is assessed for a fronthaul network based on intermediate frequency, DSP-assisted, and
digital radio-over-fiber architectures. The computed CAPEX and OPEX values show that carrier aggregation is
crucial for the cost performance and, depending on the number of aggregated carriers, one solution may be more
favorable compared to the others, and vice versa. In addition, optical fiber leasing gives a significant increase in
OPEX comparing to the case in which the optical infrastructure has to be deployed from scratch. Contrary, the
fiber deployment in a greenfield scenario has a minor impact on overall CAPEX regardless to the number of
aggregated carriers and the type of RoF architecture considered.
ACKNOWLEDGEMENTS
This work was supported by the EU H2020-MSCA-IF-2017 project “5G-DRIVE” (gr. no. 797061), the Swedish
SRA ICT TNG project “SCENE”, the Swedish Research Council project “PHASE” (2016-04510), and the
Celtic-Plus sub-project SENDATE-EXTEND funded by VINNOVA – the Swedish Governmental Agency for
Innovation Systems.
REFERENCES
[1] F. Musumeci et al.: Dynamic placement of baseband processing in 5G WDM-based aggregation networks,
in Proc. OFC, Los Angeles, USA, May 2017, paper M2G.4.
[2] Cisco visual networking index: global mobile data traffic forecast update, 2016-2021, White Paper.
[3] M. Sung et al.: Demonstration of IFoF based 5G mobile fronthaul in 28 GHz millimeter wave testbed
supporting giga-bit mobile services, in Proc. OFC, Los Angeles, USA, May 2017, paper W1C.5.
4
ICTON 2018 We.C2.5
[4] M. Fiorani et al.: Modeling energy performance of C-RAN with optical transport in 5G network scenarios,
J. Opt. Commun. Netw., vol. 8, no.11, pp. B21-B34, Nov. 2016.
[5] L. Zhang et al.: K-means clustering based multi-dimensional quantization scheme for digital mobile
fronthaul, in Proc. OFC, San Diego, USA, March 2018, paper Th2A.51.
[6] S. Ishimura et al.: 1.032-Tb/s CPRI-equivalent data rate transmission using IF-over-fiber system for high
capacity mobile fronthaul links, in Proc. ECOC, Gothenburg, Sweden, Sep. 2017, PDP.
[7] H. Zeng et al.: Real-time demonstration of CPRI-compatible efficient mobile fronthaul using FPGA,
IEEE/OSA JLT, vol. 35, no. 6, pp. 1241-1247, 2017.
[8] X. Liu et al.: Efficient mobile fronthaul via DSP-based channel aggregation, IEEE/OSA JLT, vol. 34, no. 6,
pp. 1556-1564, Mar. 2016.
[9] S.-H. Cho et al.: Cost-effective next generation mobile fronthaul architectures with multi-IF carrier
transmission scheme, in Proc. OFC, San Francisco, USA, Aug. 2014, paper Tu2B.6.
[10] A. de la Oliva et al.: An overview of the CPRI specification and its application to C-RAN-based LTE
scenarios, IEEE Commun. Mag., vol. 54, no. 2, pp. 152-159, Feb. 2015.
[11] CPRI specification, 2018, http://www.cpri.info/
[12] OBSAI specification, 2018, http://www.obsai.com/
[13] T. Pfeiffer: Next generation mobile fronthaul and midhaul architectures, J. Opt. Commun. Netw., vol. 7,
no. 11, pp. B38-B45, Nov. 2015.
[14] L. Giorgi et al.: Subcarrier multiplexing RF plans for analog radio over fiber n heterogeneous networks,
IEEE/OSA JLT, vol. 34, no. 16, pp. 3859-3866, Aug. 2016.
[15] F. Farias et al.: Cost- and energy-efficient backhaul options for heterogeneous mobile network
deployments, Photonic Netw. Commun., vol. 32, no. 3, pp. 422-437, Dec. 2016.
Table 1. Cost of the MFH transport network components and other parameters used for the CAPEX and
OPEX estimation (1 CU = 2500 USD)
Installation MTTR, Consumption,
Component Cost (CU) ANF
time, (min) (min) (W)
10G Tx 1 - 60 1000 1.5
D-RoF
Line card 1.6 30+10 per port 120 4000 285
LO 0.0048 per 1 per port 1 per port 2.76 5
Mixer IF branch 1 per port 1 per port 2.18 -
IF-A-RoF
Power
(expression) 1 per port 1 per port 1000 -
combiner
DSP-A- FPGA 3.20 5 per port 120 50 25
RoF DAC/ADC 0.08 1 per port 5 4 0.20
AWG 1 30+1 per port 60 200 -
WSS 1.6 per port 30+1 per port 60 200 15.3 per port
Splicing
CU per fiber time (min) Other related parameters and
No. of fibers
per km × no. of their values
workers
1.37×10-5 per
2 0.0696 30×1 Unavailability
km
Transport 8 0.0456 40×1 Troubleshooting 240 min
network 0.08 per
Optical fiber 12 0.0328 55×1 Leasing cost
strand per km
0.0068 per
24 0.02 90×1 Pumping cost
strand per km
48 0.0132 120×1 Trenching cost 52 per km
72 0.0132 220×3 Ducts per trench >3
96 0.0112 255×4 fibers per duct >100
144 0.0112 335×4 EDFA price 4
192 0.01 - - -
Urban speed 20 km/h
Energy cost 4×10-5 CU per kWh
Other
Technician salary 0.08 CU per hour
Floor space rent 0.116 CU per m2 per year