TRANSACTIONS ON EMERGING TELECOMMUNICATIONS TECHNOLOGIES
Trans. Emerging Tel. Tech. 2016; 00:1–8
DOI: 10.1002/ett

RESEARCH ARTICLE

5G-XHaul: A Converged Optical and Wireless Solution for 5G
Transport Networks
´
Jesus Gutierrez1∗ , Nebojsa Maletic1 , Daniel Camps-Mur2 , Eduard Garc´a2,3 , Ignacio Berberana4 ,
´
ı
Markos Anastasopoulos5 , Anna Tzanakaki5 , Vaia Kalokidou5 , Paris Flegkas6 , Dimitris Syrivelis6 ,
Thanasis Korakis6 , Peter Legg7 , Dusan Markovic8 , George Limperopoulos9 , Jens Bartelt10 , Jay Kant
Chaudhary10 , Michael Grieger11 , Nikola Vucic12 , Jim Zou13 and Eckhard Grass1,14
1

´
IHP GmbH, Germany, 2 i2CAT Foundation, Spain, 3 Technical University of Catalonia, Spain, 4 Telefonica I+D, Spain, 5 University
of Bristol, UK, 6 University of Thessaly, Greece, 7 Blu Wireless Technology, UK, 8 TES Electronic Solutions, Germany, 9 COSMOTE
- Mobile Telecommunications S.A., Greece, 10 Technical University of Dresden, Germany, 11 Airrays GmbH, Germany, 12 Huawei
Technologies Duesseldorf, Germany, 13 ADVA Optical Networking, Germany, 14 Humboldt-Universitt zu Berlin, Germany

ABSTRACT
The common European ICT sector vision for 5G is that it should leverage on the strengths of both optical and wireless
technologies. In the 5G context, a wide spectra of radio access technologies – such as millimetre wave transmission,
massive MIMO, and new waveforms – demand for high capacity, highly flexible and convergent transport networks. As
the requirements imposed on future 5G networks rise, so do the challenges in the transport network. Hence, 5G-XHaul
proposes a converged optical and wireless transport network solution with a unified control plane based on software
defined networking. This solution is able to support the flexible backhaul and fronthaul – X-Haul – options required to
tackle the future challenges imposed by 5G radio access technologies. 5G-XHaul studies the trade-offs involving fully
or partially converged backhaul and fronthaul functions, with the aim of maximising the associated sharing benefits,
improving efficiency in resource utilisation, and providing measurable benefits in terms of overall cost, scalability and
sustainability. Copyright c 2016 John Wiley & Sons, Ltd.
∗

Correspondence

´
Jesus Gutierrez, IHP GmbH, Im Technologiepark 25, 15236, Frankfurt (Oder), Germany.
´
E-mail: teran@ihp-microelectronics.com

1. INTRODUCTION
Future wireless access will extend beyond people,
supporting connectivity for any element that may benefit
from being connected, thus, facilitating concepts like the
Internet of Things (IoT) and machine-to-machine (M2M)
communication. The key motivation for 5G is to provide
ubiquitous, high-speed, high-quality wireless broadband
coverage to meet societal and industrial needs beyond
2020. To guarantee that networks will be able to cope with
the heterogeneous landscape of future services, different
fora like the 5G Infrastructure Public Private Partnership
(5G-PPP), the ITU Radiocommunication Sector (ITU-R),
and the Next Generation Mobile Networks (NGMN) are
working on the definition of performance targets for 5G
systems [1].
The considered use cases for 5G [2] [3] are very
diverse in terms of capacity and quality of service (QoS)
requirements, such as latency, jitter, bit error rate, and
device density. This heterogeneity will lead to multiple

Copyright c 2016 John Wiley & Sons, Ltd.
Prepared using ettauth.cls [Version: 2012/06/19 v2.10]

types of radio access technologies (RATs) coexisting in the
context of 5G, and that different types of services will be
addressed by different RATs.
The current network architecture uses a relatively monolithic network and transport framework to accommodate
a variety of services. The Radio Access Network (RAN)
in 5G will become more heterogeneous and will require
flexible topologies and improved transport network performance. Providing a dedicated network for each of these
technologies or applications is very complex and not cost
effective. A possible solution to this problem would be
to use network slicing – a novel architectural concept –
to support end-to-end (e2e) network slices in the RAN,
mobile fronthaul(FH)/backhaul(BH), and core network
(CN). In order to support network slicing in the mobile
transport network, dynamic network resource management
and sharing will be necessary, which requires an integrated
mobile transport network, including the mobile FH and
BH.

1

´
J. Gutierrez et al.

In view of this, the 5G-PPP project 5G-XHaul aims
at developing a transport network able to provide
transport services on demand, seen as a generic data
forwarding public resource, where every terminal can
directly communicate with any other. This involves the
development of a suitable flexible and dynamic data plane,
based on the integration of optical and wireless network
technologies, that is controlled by a unified control plane.
Current transport networks, especially those for FH,
have to fulfill stringent requirements in terms of data rates,
latency and/or synchronisation. The introduction of new
RAN technologies is expected to have a major impact on
the transport network, including:
• The optimal functional split of processing functions
executed locally at the central unit (CU), or
remotely at the Remote Unit (RU) – also known
as Baseband unit (BBU) and Remote Radio Head
(RRH), respectively – in the Common Public Radio
Interface (CPRI). This optimal split represents a
key architectural issue, since it relaxes the stringent
requirements for transport capacity, delay and
synchronisation. The selection of this split depends
on the transport network characteristics, the
network topology and scale, the type and volume
of services to be supported. These functional splits
will be further discussed in Section 3.1.1.
• The introduction of higher carrier frequencies in
the 5G RAN, such as centimetre wave (cmWave)
and millimetre wave (mmWave) in licensed and
unlicensed bands which will enable higher data
bandwidths and capacity at the expense of higher
path losses. The wide bandwidths on the air
interface will also require higher capacities in
the transport network and potentially an even
stricter synchronisation. The increase in the number
of antennas, including Massive MIMO, will also
require higher capacities in the transport network.
• The plurality of RAN technologies under discussion
in the context of 5G. Examples are higher
order modulation schemes, new waveforms and
technologies for the support of improved latency
and resilience, and a large variety of use cases
and machine type communications. These new use
cases will have to be reflected in the transport
network as well, i.e. by providing low latency or
high reliability.
In this context, convergence of wireless and optical
communications appears to be a strong candidate to fulfill
the stringent requirements imposed by the RAN and new
RATs. The main goal of the 5G-XHaul project is to develop
a novel converged optical/wireless architecture and a
suitable network management solution for supporting
mobile scenarios.
The remainder of this paper is organised as follows.
Section 2 describes the different layers of the 5G-XHaul
architecture. Section 3 presents the key benefits of the
proposed converged transport network along with the
2

enabling technologies, interfaces and control management
methods. Section 4 highlights the 5G-XHaul intentions
to contribute to standardisation bodies and sketch the
expected impact. Finally, the conclusions are presented in
Section 5.

2. FUNCTIONAL ARCHITECTURE
5G-XHaul proposes a converged optical/wireless 5G
network infrastructure interconnecting computational
resources with fixed and mobile users to support both
operational network – Cloud RAN (C-RAN) – and
end-user computational services, adopting the concept of
cloud computing. A layered architecture, inspired by the
ETSI Network Function Virtualisation (NFV) standard
[4], and the SDN reference architecture [5], are considered
to effectively and efficiently provision both end-user and
operational services over the proposed infrastructure.
5G-XHaul will adopt a holistic view considering jointly
mobile FH and BH functions to ensure appropriate
allocation of the required resources across all domains.
Its objective is twofold: to minimise the operational
expenditure of the FH in terms of power consumption
and capital expenditure under the associated strict delay
constraints, and to minimise the total e2e service delay in
the BH.
Examples of features that enable these benefits include
the option to virtualise the separate control plane,
using NFV and deploying Virtual Network Functions
(VNFs). These are controlled by the SDN controller, to
allow on demand resource allocation, able to support
dynamically changing workloads [6]. SDN network
elements can be treated as VNFs, since they can be
implemented as software running on general-purpose
platforms in virtualised environments. Both SDN and
non-SDN models can be supported by SDN network
elements. On the other hand, network applications can
include SDN controller functions, or interact with SDN
controllers and can themselves provide VNFs. Service
Chaining, supporting orchestrated service provisioning
over heterogeneous environments, is considered to be one
possible network application. Network elements controlled
by SDN controllers can provide either VNFs or Physical
Network Functions.
Taking advantage of the SDN concept and the
benefits of cross-technology virtualisation, 5G-XHaul
proposes an overarching layered architecture able to
efficiently and effectively support 5G services [7],
presented in Figure 1. The 5G-XHaul infrastructure, in
accordance to the 5G vision, exhibits a great degree
of heterogeneity in terms of technologies. To address
the challenge of managing and operating this type
of complex heterogeneous infrastructure in an efficient
manner, 5G-XHaul proposes the adoption of SDN and
NFV that will be seamlessly integrated. In SDN, the
control plane is decoupled from the data plane and
is managed by a logically centralised controller that
Trans. Emerging Tel. Tech. 2016; 00:1–8 c 2016 John Wiley & Sons, Ltd.
DOI: 10.1002/ett
Prepared using ettauth.cls

· high network bandwidth
· Increased BBU sharing

vBBU
(lower layers)

Virtual
Wireless

Virtual BH

End
Point A

forwarding

End
Point B

forwarding
Virtual
Optical

Virtual
Wireless

Virtual
Optical

VM

(4)

EM

Element Management (EM)

EM

SDN
Controller

VNF

SDN
Controller

VNF

(2)
Cycle Prefix
& FFT
RF to
Baseband

· low network bandwidth
· Limited BBU sharing

Inf Management

Receive
processing
Resource
demapping

(1)

Element Management
(EM)

Control

Decoding

(3)
Traditional RAN

Virtual FH vBBU
(upper layers)

RU

(5)

Managed PHY Inf

Cloud RAN

MAC

Management & Service
Orchestration

´
J. Gutierrez et al.

Virtual
Network

Processing

Virtual
Network

Virtualisation

Virtualisation

mmWave
Network
Controller

Processing
Physical
Network

Physical
Network

LTE Network
Controller

Virtualisation

Compute
Controller

WDM PON
Controller

Backhaul

RRH
RRH RRH

Virtualisation

TSON
Controller

Fronthaul
vBBU

vBBU
VM

Drivers

Data Centers
Wireless Access/Transport

Compute
Controller

Internet

VM
Optical Transport

Data Centers

Figure 1. The overarching architecture supporting functional split processing [7].

has a holistic view of the network [8]. At the same
time, NFV enables the execution of network functions
on commodity hardware (general-purpose servers) by
leveraging software virtualisation techniques [8]. Through
joint SDN and NFV consideration, significant benefits can
be achieved. These are the flexible, dynamic and efficient
use of the infrastructure resources, simplification of the
infrastructure and its management, increased scalability
and sustainability as well as provisioning of orchestrated
e2e services.
This layered architecture includes a Physical Infrastructure (PI) layer that comprises a converged optical/wireless
transport network able to interconnect RUs, end-users and
computing resources. The wireless access part includes
dynamically programmable, high capacity, low latency,
point-to-multipoint (P2MP) mmWave transceivers, cooperating with Sub-6 GHz systems. The optical transport
relies on a hybrid Wavelength Division Multiplexing
(WDM) passive optical network (PON) and active optical
network solutions able to handle elastic sub-wavelength
switching granularities.
The PI will be managed by the Infrastructure Management layer (IML) responsible for the management of
the different technology domains and the creation of virtual infrastructure and PI slices comprising heterogeneous
resources. These infrastructure slices will enable virtual
operator models providing both FH and BH services. The
IML will communicate with the various network and compute controllers that are responsible for retrieving information and communicating with the individual domains.
Once the information has been collected, the resources are
abstracted and virtualised. From the architectural and functional perspective, the infrastructure management layer

Trans. Emerging Tel. Tech. 2016; 00:1–8 c 2016 John Wiley & Sons, Ltd.
DOI: 10.1002/ett
Prepared using ettauth.cls

addresses all virtualisation associated functions as well as
the virtual resource management functions. Management
of traditional non-virtualised PIs can be also supported.
Cross-domain orchestration of the virtual and physical
infrastructures, created and exposed by the IML to the
higher layers, will be carried out by a control layer.
This layer needs to implement converged control and
management procedures for dynamic and automated
provisioning of e2e connectivity services according to
specific QoS considerations.
Finally, a Management and Service Orchestration
layer responsible for the converged orchestration of
end user services is needed. This can be used for
the composition and delivery of multi-tenant chains
of virtualised network functions. In addition, it can
perform resource orchestration through NFV resources
across multiple Virtual Infrastructure Managers (VIMs)
and include lifecycle management of Network Services,
supporting Network Service Orchestration functions.
Service Orchestration layer related activities are out of
scope of the 5G-XHaul project.

3. 5G-XHAUL DEVELOPMENT FOCUS
5G-XHaul is defining a converged wireless and optical
transport network offering a converged datapath to an
SDN-based control plane.
3.1. 5G-XHaul Data Plane Overview
One of the first steps to understand the nature of the
envisioned 5G transport network has been so far to work
on analysing the diverse functional splits from the RAN
perspective. This means identifying the optimal ways to
3

3.1.1. Functional splits and Transport Classes
As mentioned in the introduction, new 5G RAN
technologies such as mmWave or Massive MIMO will
greatly increase the requirements on FH networks.
However, a full centralisation of baseband processing
might not be beneficial in all scenarios or for all
air interface technologies. Hence, different functional
splits (i.e., the allocation of baseband functionality to
either RUs or CUs [9][10]) could be employed in 5G
networks to enable a high degree of centralisation where
it promises benefits. In other scenarios, to reduce costs,
the transport network could have much less stringent
requirements. These functional splits blur the difference
between traditional FH and BH network, which motivates
one key concept of 5G-XHaul, the converged transport
network. The converged transport network unifies the FH
and BH equipment and, hence, reduce deployment as
well as operational and maintenance cost. As each of the
functional splits imposes a different set of requirements on
the transport network, 5G-XHaul has introduced the notion
of transport classes (TCs) [11][12]. These TCs would be
used to identify traffic corresponding to different sets of
requirements in order to facilitate, e.g., a prioritisation of
processing with a low overhead in the transport network.
5G-XHaul focuses on three functional splits, illustrated
in Figure 2. Split A basically corresponds to a full
centralisation as utilised by the CPRI standard, which
is used in most of today’s C-RAN deployments. The
only difference to CPRI is that antenna mapping is
performed decentralised, to enable large antenna arrays
for beamforming (BF) without having to transport a
separate stream for each antenna element. Otherwise, all
processing is centralised and I/Q samples are forwarded
between CU and RU. However, this comes at the price
of the aforementioned stringent requirements in terms
of data rate, latency, and synchronisation. One dominant
disadvantage of CPRI is that the FH data rate is constant
and independent of the actual load of the cell.
Split B utilises decentralised resource mapping.
Thereby, most PHY processing can still be performed
centrally, but only actually utilised resource blocks need
to be forwarded on the transport network. This can greatly
reduce the capacity to be deployed in the aggregation
segment of transport networks [11][12], and is mostly in
line with the Next Generation Fronthaul Interface (NGFI)
currently under investigation [13][14].
Splits A and B require synchronisation in order to
align samples, hence a transport class was introduced
for synchronisation services as mentioned above. As the
most stringent delay requirement in CPRI arises from
4

Backhaul

Split C

Modulation
mapper

Scrambling

Modulation
mapper

Resource
OFDM
element
signal
mapper generation
Resource
OFDM
element
signal
mapper generation

Split B

Split A

Antenna
mapping

Scrambling

Precoding

S1

Layer
mapper

PDCP
Upper MAC
+ RLC

split a RAN processing chain in a Remote Unit (RU)
and a Centralised Unit (CU), and studying the trade-offs
of placing these RUs/CUs in distributed or centralised
locations [9]. The 5G-XHaul transport network consists of
an heterogeneous data plane involving advanced wireless
and optical network technologies and designed for coping
with the requirements imposed by the RAN.

Lower MAC +
HARQ
CRC & Coding

´
J. Gutierrez et al.

DAC

Radio 1

DAC

Radio N

Fronthaul

Figure 2. Functional splits options investigated by 5G-XHaul.

the hybrid automated repeat request (HARQ) scheme,
split C decentralises all processing up to and including
HARQ. Thereby, the HARQ delay requirement for the
FH network is removed, allowing for greater distances
between RUs and CUs or allowing for additional latency
contributions by transport nodes such as routers or
switches. However, this disables the option for centralised
PHY-layer processing, and hence centralisation gains
can only come from technologies such as centralised
scheduling or overall network management. In terms
of overall transport requirements this split is similar to
traditional BH. As a final remark, it is worth noting that
a particular split is better suited for a certain scenario than
for all, but it is expected that multiple splits will coexist
within the same 5G network [11].
As an output of this analysis, 5G-XHaul is proposing
a set of TCs as a foundation of 5G transport. These TCs
have been identified for: a) TC0: synchronisation traffic,
b) TC1: transport of I/Q samples/tactile internet traffic, c)
TC2: packet-based FH with statistical multiplexing, and d)
TC3: traditional BH services.
3.1.2. Heterogeneous Technologies
In order to fulfil the requirements associated to the
TCs introduced in Section 3.1.1, 5G-XHaul focuses on
developing novel programmability features on data plane
technologies. For example: 1) Flexibly allocating time
and frequency resources in high capacity, flexible optical
transport, 2) Dynamically turning off WDM-PON Optical
Network Unit (ONU) transceivers to save energy, or 3) Reconfiguring beam directions in mmWave nodes to enable
P2MP in the Small Cell wireless BH. Some of these
enabling technologies are briefly introduced below.
Optical Technologies: TSON, WDM-PON 5G optical
transport networks able to support converged FH and BH
services need to provide a high degree of flexibility both
in time and frequency domains. Adaptability, scalability,
energy efficiency and resilience are other requirements of
key importance.
A key enabler supporting the feasibility of this approach
is the adoption of a high capacity, flexible optical transport
comprising a combination of passive and active solutions.
PON solutions will be based on WDM-PONs, while active
solutions will adopt more flexible and dynamic WDM
technologies such as the Time-Shared Optical Network
(TSON).
A pictorial representation of the 5G-XHaul optical
transport is shown in Figure 3. TSON has already been
deployed at a European level network testbed to efficiently
and effectively demonstrate advanced backhauling capabilities [15]. It is designed and implemented as a novel
Trans. Emerging Tel. Tech. 2016; 00:1–8 c 2016 John Wiley & Sons, Ltd.
DOI: 10.1002/ett
Prepared using ettauth.cls

´
J. Gutierrez et al.

Small Cell

Elastic TSON frame
Frequenc
y

RRH

PON

Macro-cell

time

BBU pool

Light IP flow
Large CPRI flow

Figure 3. 5G-XHaul Optical Transport
ONU

OLT

5G-XHaul
Ethernet
switch

1
2

...

MUX

..
...

Rx Array
1
2

...
n

C

Single mode Fiber
20~40 km

n

..
...

L

ONU 1
L

L

C

C

T-LD
Rx

ONU n
T-LD

DEMUX

BBU

Cross-connect

…

RN

Tx Array

Cyclic AWG

OLT

BBU

Rx

Figure 4. 5G-XHaul WDM-PON Solution.

frame-based, time multiplexing metro network solution,
offering dynamic connectivity with fine bandwidth granularity [16].
The TSON framework offers a very flexible optical
platform that supports sub-wavelength switching of 10 G
Ethernet (10GE) frames, flexible length, programmable
frames varying from 64 ns to 25.6 us and variable bit rates
spanning from of 30 Mbps up to 6 Gbps. As TSON’s
operational characteristics can be dynamically modified,
varying service related requirements can be supported.
Parameters that can be dynamically modified in the current
TSON version include bit-rate, burst length, connectivity
type and QoS. Extensions planned to be developed in
the framework of 5G-XHaul will include also elastic
bandwidth allocation and more stringent synchronisation
capabilities.
In combination with the TSON technology described
above, a WDM-PON solution is designed to deliver the
point-to-point (P2P) connectivity between regional data
centers that host BBUs and the RUs, i.e. the RAN network,
and also able to provide BH connectivity from the BBU
to an Ethernet switch, as shown in Figure 4. Each WDM
channel is currently able to achieve bit rates up to 10 Gbps,
over a transmission distance of 20 to 40 km.
A wide variety of flows will be transported over the 5GXHaul infrastructure, which span from small control flows
to medium IP BH flows and large Cloud-RAN FH flows.
Therefore, flexible addressing of individual transport flows
between different BBUs and ONUs is required in the
proposed system. In addition, flexible RAN networks
incorporating flexible centralisation strategies also need to
be orchestrated by the 5G-XHaul control plane. In order
to realise such a dynamicity in the RAN deployment,
an electrical or optical cross-connection functionality
between the BBU pool and the transceiver array of the
WDM-PON system is provided, such that the traffic flow
from each BBU can be flexibly assigned to any desired
RRH.

Trans. Emerging Tel. Tech. 2016; 00:1–8 c 2016 John Wiley & Sons, Ltd.
DOI: 10.1002/ett
Prepared using ettauth.cls

Wireless Technologies: mmWave, Sub-6 GHz Optical
fibre links are widely available in urban environments
but still their deployment is not fast enough to match
the sheer number of small cells required to serve the
upcoming demand in mobile data. The high deployment
cost of optical fibre will be alleviated in 5G-XHaul with
the use of mmWave communications for BH purposes. The
goal is to deliver high capacities – provide the potential
gigabit-per-second traffic demand for wireless transport
– using mmWave links operating in the 60 GHz band.
In this context, it is important to exploit novel radio
signal processing methods in the PHY layer, demonstrating
that Line-of-Sight and Non-Line-of-Sight performance for
mmWave systems can be enhanced.
In very dense RANs, multi-hop wireless topologies for
the BH are required where wireless links will transport
the aggregated traffic of several BSs. The idea is to
overcome the limitation imposed by static P2P links with
dynamic P2MP mmWave connectivity models. 5G-XHaul
will develop BF chipsets and design techniques to allow
dynamic and autonomic reconfiguration of the mmWave
BH, enabling the network management to control the
BF weights used at individual mmWave nodes. These
techniques will also allow the implementation of different
P2MP schemes over several mmWave nodes.
Antennas for P2MP wireless BH at 60 GHz have to
be highly directive, with high gain enabling efficient BF
and easy integration with the BF chipset. Planar (printed)
antenna arrays are considered due to their small form
factor, good reproducibility and low cost. Phased antenna
arrays with patches or dipoles as basic antenna elements
are investigated, with the target to enable scanning in
both horizontal and vertical direction. The advantages and
limitations of both types are studied and the one with
better properties (gain and complexity) will be chosen for
implementation.
Sub-6 GHz technologies, such as IEEE 802.11ac, will
be considered as one of the key building blocks of the 5GXHaul wireless BH. The network-wide performance of the
Sub-6 GHz BH will be improved through the intelligent
orchestration of the individual BH elements. In 5G-XHaul,
the Sub-6 GHz wireless BH will be tightly coupled with
the mmWave BH in order to implement an intelligent
transport network. For example, traditional IP BH flows
could be transported through the Sub-6 GHz BH while
the mmWave BH is devoted to transport heavy CloudRAN FH flows. Indeed, 5G-XHaul will consider deeper
levels of integration where the Sub-6 GHz BH will be
used to transport control plane information to configure
the mmWave transport. It is worth mentioning that Sub1 GHz transmissions can also be integrated as part of
the 5G-XHaul portfolio where bandwidth requirements are
relaxed. Sub-1 GHz technologies, such as the new IEEE
802.11ah, provide longer and more reliable wireless links
(up to 10x the range of IEEE 802.11ac), at the cost of
reducing capacity.

5

´
J. Gutierrez et al.

ETN

3.2. 5G-XHaul Control Plane Overview
The 5G-XHaul control plane allows tenants – users
of the 5G-XHaul transport system –, which receive
connectivity services over the transport network as, e.g.
virtual 5G network operators – to define a transport slice
specifying how a set of distributed physical or tenantspecific virtual functions are connected through the 5GXHaul infrastructure. In addition, the 5G-XHaul control
plane features a North Bound Interface (NBI) that allows
tenants to independently control the functions in their
transport slices.
In order to support concurrent per-tenant slices in
an efficient and scalable way, 5G-XHaul decouples
the per-tenant state from the transport-specific state.
Following architectural approaches that have been proven
to scale in data center environments [17], 5G-XHaul
keeps the per-slice state in an overlay composed of
edge devices, known as Edge Transport Nodes (ETNs),
and illustrated in Figure 5. Consequently, the underlay
transport infrastructure focuses on providing connectivity
between ETNs in a slice agnostic manner. This approach
does not only allow 5G-XHaul to scale, but it also allows
to integrate diverse transport technologies in the transport
underlay, which is key to enable an easy migration path
from legacy transport networks to a 5G-XHaul transport.
However, although being able to function with legacy
transport underlays, 5G-XHaul features native support for
a transport underlay based on an Ethernet-like datapath
and an SDN control plane. Specifically, to improve
scalability and support for heterogeneous technologies,
5G-XHaul partitions the transport infrastructure in a
set of control plane areas, depicted as Tier-0 Areas
(T0-Area) in Figure 5, where transport devices within
each area are controlled by a corresponding Tier-0
Controller (T0-Ctrlr). Notice in Figure 5 how a T0-Area
consists of ETNs that connect per-tenant functions to
the 5G-XHaul infrastructure, Inter-Area Transport Nodes
(IATNs), which interconnect T0-Areas, and Transport
Nodes (TNs) connecting ETNs and IATNs within an area.
ETNs, TNs and IATNs are controlled by the corresponding
T0-Ctrlr using specific southbound interfaces (SBIs). The
control plane is organised hierarchically, whereby T0
controllers feature a NBI towards Tier 1 controllers (T1Ctrlr), which operate at a higher abstraction level, and,
among other functions, support forwarding at the inter-area
level. At the highest layer of the hierarchy, we can find the
top level controller offering the 5G-XHaul NBI to services
and other components of the 5G system.
Edge Transport Nodes (ETNs) ETNs maintain perslice state providing 5G-XHaul tenants the required
abstraction to operate on their slices. For this purpose, an
ETN features the internal components shown in Figure 6.
On the control plane, an ETN instantiates a logical
datapath for each tenant having functions connected to
that ETN. Logical datapaths receive high level control
policies from the tenant’s own control plane, and push
those policies to a local SDN controller in the ETN. The
6

Logical DPTenant 1

External Per
tenant Hierarchical
Cache

Slice 1

Logical DPTenant N

Local Ctrlr

Slice 2

Per-tenant FIB

ETN

TN-adaptation

IA-TN
TN

TN

TN

TN

T0-Area 1

TN
T0-Area 2

Figure 6. Detailed view of an ETN.

local controller in the ETN implements the corresponding
per-tenant state by accordingly placing the rules on each
logical datapath. Introducing multi-tenancy support at this
last controller layer, which runs on the actual forwarding
element, leaves intact the underlying performance-centric
multi-stage FIB architecture design that performs the
actual packet forwarding. Hence, a critical aspect in the
design of the ETN is the datapath delay introduced by the
per-tenant Forwarding Information Base (FIB). In order to
minimise this delay, tenant-specific rules are often evicted,
resulting in a small number of simultaneous rules hosted
in the FIB. Consequently, a hierarchical structure of rule
caches is used to scale up to a large number of rules coming
from different tenants. Finally, the per-tenant FIB matches
tenant-specific rules and inserts packets into transportspecific tunnels. A Transport Network adaptation function
is included in the ETN that pushes the corresponding
transport header before injecting the packets into the
transport network. In the case of a native 5G-XHaul
transport, the TN Adaptation function injects an outer
Ethernet header like the one used in Provider Backbone
Bridging (PBB) [18].
Transport Nodes (TNs) TNs provide connectivity
between ETNs and IATNs within a given T0-Area by
means of transport tunnels. Thus, TNs forward incoming
packets based on the corresponding tunnel identifiers,
which, in the case of the native 5G-XHaul transport
underlay, are defined based on a destination MAC address,
a transport specific VLAN, and the I-SID field identifying
a transport slice. The proposed Ethernet-based datapath is
designed to support a variety of forwarding paradigms such
as pinned down tunnels and source routing [19].
Inter Area Transport Nodes (IATNs) IATNs connect
together multiple T0-Areas, and, as depicted in Figure 6,
are functionally composed of a set of TNs, one for each
of the areas connected by that IATN, joined together by
an IATN specific control and data plane functions. In the
control plane, the IATN function conveys to the 5G-XHaul
control plane the T0-Areas that can be reached through
this IATN. In the data plane, the IATN translates transport
tunnel headers when adjacent areas connected by the IATN
feature different transport technologies.
3.2.1. 5G-XHaul QoS and TE framework
5G-XHaul considers a two level QoS framework. At
the underlay transport level, the four core transport classes
Trans. Emerging Tel. Tech. 2016; 00:1–8 c 2016 John Wiley & Sons, Ltd.
DOI: 10.1002/ett
Prepared using ettauth.cls

´
J. Gutierrez et al.

NBI
T2-Ctrlr
T1-Ctrlr#2
T1-Ctrlr#1
T0-Ctrlr#1
T0-Ctrlr#2

T0-Ctrlr#3

T0-Area#1
T0-Ctrlr#5
T0-Area#2

T0-Area#3
T0-Area#4

T0-Area#5
Edge Transport Node (ETN)
Transport Node (TN)
Inter-Area Transport Node (IATN)
(v)Function from tenant 1 (slice 1)
(v)Function from tenant 2 (slice 2)

Figure 5. 5G-XHaul Control Plane architecture.

introduced in Section 3, are supported between ETNs
and IATNs within each area through the instantiation
of different transport tunnels for each class, which, for
example, allows to consider minimum delay paths for
TC0 and TC1. In addition, within an ETN, and possibly
all nodes even including TNs, QoS functionalities are
included such as Priority or Weighted fair queueing (WFQ)
schedulers that allow to prioritise traffic injected at each
transport tunnel interface at the inter-slice and intra-slice
level.
Traffic Engineering (TE) functions at different
timescales are also supported by the hierarchical 5GXHaul control plane. Local and T0 controllers make more
dynamic TE decisions at short timescales while higher
tier controllers implement TE functions at longer epochs.
Examples of TE functions supported include P2P and
P2MP tunnel setup, load balancing between alternative
(equal cost) transport tunnels, resilience through the
establishment of backup tunnels, traffic policing and
shaping.
3.2.2. 5G-XHaul RAN-Transport Interface
A core function of the 5G-XHaul control plane is
the ability to interface with 4G or 5G RANs in order
to estimate spatio-temporal demand variations in the
RAN and allocate transport resources accordingly. For
this purpose, 5G-XHaul is developing several interfaces
that allow the exchange of information between transport
and RAN at different time scales. First, an interface
is considered between the upper tiers of the 5G-XHaul
control plane and the Mobility Management Entity – or
its 5G equivalent – of the Mobile Network in order to
monitor varying traffic demands at a macro scale (e.g. by
means of anonymised traces, such as number of users per
tracking area, or estimated weekly patterns), which aid
planning decisions in the transport network. Second, an
interface between a RAN controller and the T0 controller
is considered at shorter time scales that can be used to
Trans. Emerging Tel. Tech. 2016; 00:1–8 c 2016 John Wiley & Sons, Ltd.
DOI: 10.1002/ett
Prepared using ettauth.cls

optimise RAN transport coordination for handover or load
balancing purposes. Finally, a tightly coupled interface
between the RAN scheduler and the 5G-XHaul ETN is
considered to allow the RAN scheduler to quickly change
the point of attachment of a flow, which is a critical feature
in cell-less architectures considered, e.g. in mmWave
RANs [20].

4. 5G-XHAUL CONTRIBUTION TO
STANDARDS
5G-XHaul is well timed to help shape the transport
network definition within 5G standardisation. Early results
can be fed into the study phase of 5G definition, which
is currently underway in 3GPP, whilst partners can push
later findings into the normative specifications. The target
date for the first standards release by 3GPP is late 2020.
At the same time, work is ongoing to specify transport
technologies that are not necessarily targeted solely at
5G, for example in optical and microwave FH/BH, so the
project aims to contribute here also.
This project has identified a number of domains and
standardisation bodies therein where contributions can be
made. The functional architecture of the 5G RAN and CN
together with interfaces will be defined in 3GPP, guided
by NGMN and approved by the ITU-T. The sharing of
transport links for FH and BH, and the sharing of radio
links for access and BH, are examples of aspects where
the project can contribute. As described above, the project
will develop and demonstrate both wireless and optical
transport technologies, with outputs feeding into European
spectrum regulation, the latest IEEE 802.11 wireless
access (ay and ax), and WDM-PON developments in ITUT SG15 and FSAN. Finally, the 5G-XHaul control plane
exploits SDN, consequently OpenFlow extensions will be
submitted to the ONF, and extensions developed for the
OpenDayLight SDN controller used in the demonstration
will be fed back to the community.
7

´
J. Gutierrez et al.

5. CONCLUSIONS
5G infrastructures will not rely on a single RAT, but
will support heterogeneous technologies, including the
traditional 2G, 3G, 4G, as well as new 5G air interfaces,
WiFi solutions, etc. This variety of RATs and the targeted
applications have very diverse QoS requirements, need
to support different traffic models and rely on different
architectures. This paper provides the overall view of the
5G-XHaul transport network architecture, which is based
on the convergence of optical and wireless technologies for
serving future 5G RANs. This convergence also applies
to FH and BH networks. The main goal of a unified
convergent transport network is the reduction of cost.
This can be achieved by using the same infrastructure
to transport data from different functional splits and of
different RATs. The network allows dynamic adaption to
different traffic patterns and usage scenarios. A coherent
management is achieved by SDN functions in conjunction
with NFV.

8.

9.

10.

11.

12.

ACKNOWLEDGEMENT
This work has received funding from the European Union’s
Horizon 2020 research and innovation programme under
grant agreement No. 671551. The European Union and
its agencies are not liable or otherwise responsible for the
contents of this document; its content reflects the view of
its authors only.

13.

14.

15.

REFERENCES
1. 5G Infrastructure PPP Association, 5G Vision - The
5G Infrastructure Public Private Partnership: the next
generation of communication networks and services,
White Paper, February 2015.
2. NGMN Alliance, NGMN 5G White Paper, White
paper, Feb. 2015. [Online]. Available: http://
ngmn.org/uploads/media/NGMN_5G_
White_Paper_V1_0.pdf. Accessed Mar. 20,
2016.
3. METIS
II
D1.1
Refined
scenarios
and
requirements, consolidated use cases, and qualitative
technoeconomic feasibility assessment, https://
metis-ii.5g-ppp.eu/wp-content/
uploads/METISII_D1.1_v1.0.pdf. Accessed
Mar. 30, 2016.
4. Network Functions Virtualisation, ETSI, http://
www.etsi.org/technologies-clusters/
technologies/nfv. Accessed Mar. 23, 2016.
5. ONF, SDN reference architecture, https://www.
opennetworking.org/. Accessed Mar. 28, 2016.
6. ETSI GS NFV-SWA 001 V1.1.1 (2014-12), Network
Functions Virtualisation (NFV); Virtual Network
Functions Architecture, 2014.
7. A. Tzanakaki et al., ”5G Infrastructures Supporting
End-User and Operational Services: The 5G-XHaul
8

16.

17.

18.

19.

20.

Architectural Perspective”, Accepted to IEEE ICC
2016, Workshop on 5G Architecture, Kuala Lumpur,
Malaysia, May 2016.
B. Han, V. Gopalakrishnan, L. Ji, S. Lee, ”Network
function virtualization: Challenges and opportunities
for innovations,” IEEE Commun. Mag., vol.53, no.2,
pp.90-97, Feb. 2015.
J. Bartelt et al., ”Fronthaul and backhaul Requirements
for Flexibly Centralized Radio Access Net-works”,
IEEE Wireless Comm. Mag., Vol. 22, no. 5, pp 105111, October 2015.
U. D¨ tsch, M. Doll, H. P. Mayer, F. Schaich,
o
J. Segel, and P. Sehier, ”Quantitative Analysis of
Split Base Station Processing and Determination of
Advantageous Architectures for LTE,” Bell Labs Tech.
J., vol. 18, no. 1, pp. 105-128, May 2013.
J. Bartelt et al., ”The Impact of 5G Radio Access
Technologies on a Converged Fronthaul and Backhaul
Network,” submitted to IEEE Wireless Comm., Feb.
2016.
Deliverable D2.1 ”Requirements Specification and
KPIs”, 5G-XHaul Project, March 2016.
C. L. I, J. Huang, Y. Yuan, S. Ma and R. Duan, ”NGFI,
the xHaul,” 2015 IEEE Globecom Workshops (GC
Wkshps), San Diego, CA, 2015, pp. 1-6.
China Mobile NGFI White paper: China Mobile,
ALU, Nokia, ZTE, Broadcom, Intel, White Paper of
NGFI (Next Generation Fronthaul Interface) Version
1.0, October 4, 2015.
B. Rahimzadeh Rofoee, K. Katsalis, Y. Yan, Y.
Shu, T. Korakis, L. Tassiulas, A. Tzanakaki, G.
Zervas, D. Simeonidou, ”Demonstration of Servicedifferentiated Communications over Converged Optical Sub-Wavelength and LTE/WiFi Networks using
GEANT Link,” in proc. of OFC 2015, Los Angeles,
California, USA
Y. Yan et al., ”High performance and flexible FPGAbased time shared optical network (TSON) metro
node,” Opt. Exp 21, 5499-5504,2013.
T. Koponen et al., ”Network virtualization in multitenant datacenters [next-generation carrier ethernet
transport technologies, ” In 11th USENIX Symposium
on Networked Systems Design and Implementation
(NSDI 2014) (pp. 203-216).
D. Fedyk and D. Allan. ”Ethernet data plane evolution
for provider networks”, IEEE Communications Magazine, vol. 46, no. 3, pp. 84-89, March 2008.
C. Filsfils, N. K. Nainar, C. Pignataro, J. C. Cardona,
P. Francois, ”The Segment Routing Architecture”, In
IEEE Global Communications Conference (GLOBECOM 2015), San Diego, California, USA (pp. 1-6).
R. J. Weiler, M. Peter, W. Keusgen, E. C. Strinati, A.
De Domenico, I. Filippini, A. Capone, I. Siaud, A.M.
Ulmer-Moll, A. Maltsev and T. Haustein, ”Enabling
5G backhaul and access with millimeter-waves”, In
EuCNC 2014, Bologna, Italy, June 2014 (pp. 1-5).

Trans. Emerging Tel. Tech. 2016; 00:1–8 c 2016 John Wiley & Sons, Ltd.
DOI: 10.1002/ett
Prepared using ettauth.cls