smart backhauling and fronthauling 5g networks …

IEEE Wireless Communications • October 201532 1536-1284/15/$25.00 © 2015 IEEE

Antonio de la Oliva iswith University Carlos IIIof Madrid.

Xavier Costa Pérez andJohannes Lessmann arewith NEC Europe Ltd,Germany.

Arturo Azcorra is withUniversity Carlos III ofMadrid and IMDEA Net-works.

Andrea Di Giglio is withTelecom Italia.

Fabio Cavaliere andPaola Iovanna are withEricsson Research, Italy.

Dirk Tiegelbekkers is withNokia Solutions and Net-works.

Thomas Haustein is withFraunhofer HeinrichHertz Institute.

Alain Mourad is withInterDigital Europe Ltd.

SMA RT BA C K H A U L I N G A N D FR O N T H A U L I N G

F O R 5G NE T W O R K S

THE NEED FOR A NOVEL INTEGRATEDBACKHAUL/FRONTHAUL NETWORK

According to recent predictions [1], mobile datatraffic will increase 11-fold between 2013 and2018. Fifth generation (5G) radio access network(RAN) technologies serving this mobile datatsunami will require fronthaul and backhaul

solutions between the RAN and packet core todeal with this increased traffic load. Further-more, there will be a sizeable growth in the cap-illarity of the network since traffic load increasein the 5G RAN is expected to stem from anincreased number of base stations with reducedcoverage (i.e., mobile network densification).

To support the increased density of themobile network (e.g., in terms of interferencecoordination) and achieve the required 5Gcapacity, extensive support for novel air interfacetechnologies such as cooperative multipoint(CoMP), carrier aggregation (CA), and massivemultiple-input multiple-output (MIMO) [2] willbe needed. Such technologies require processingof information from multiple base stations simul-taneously at a common centralized entity andalso tight synchronization of different radio sites.Hence, backhaul and fronthaul will have to meetmore stringent requirements not only in terms ofdata rate but also in terms of latency, jitter, andbit error rate [3].

Given that the aforementioned challengeswill be addressed by service providers in a busi-ness environment where revenues do notincrease proportionally to the data volume, cost-effective network deployment, efficient opera-tion, and evolution strategy are required. Apromising approach to address this challenge, ascan be observed in major standardization bodies,such as the European Telecommunications Stan-dards Institute (ETSI) network functions virtual-ization (NFV), is virtualization, which exploitsthe multiplexing gain of softwarized networkfunctions on top of commoditized hardware.This has led to the centralized RAN [4, 5] con-cept where cellular base station functions arehosted in cloud computing centers. Once virtual-ized, base station functions can be flexibly dis-tributed and moved across data centers,providing another degree of freedom for loadbalancing. Moreover, base station functions canbe decomposed in many different ways [6], giv-

ANTONIO DE LA OLIVA, XAVIER COSTA PÉREZ, ARTURO AZCORRA, ANDREA DI GIGLIO, FABIO CAVALIERE, DIRK TIEGELBEKKERS, JOHANNES LESSMANN, THOMAS HAUSTEIN,

ALAIN MOURAD, AND PAOLA IOVANNA

ABSTRACTThe Xhaul architecture presented in this arti-

cle is aimed at developing a 5G integrated back-haul and fronthaul transport network enablingflexible and software-defined reconfiguration ofall networking elements in a multi-tenant andservice-oriented unified management environ-ment. The Xhaul transport network vision con-sists of high-capacity switches and heterogeneoustransmission links (e.g., fiber or wireless optics,high-capacity copper, mmWave) interconnectingremote radio heads, 5G points of attachment(5GPoAs, e.g., macro- and small cells), central-ized-processing units (mini data centers), andpoints of presence of the core networks of one ormultiple service provider(s). This transport net-work shall flexibly interconnect distributed 5Gradio access and core network functions, hostedon network centralized nodes, through the imple-mentation of a control infrastructure using a uni-fied, abstract network model for control planeintegration (Xhaul Control Infrastructure, XCI);and a unified data plane encompassing innova-tive high-capacity transmission technologies andnovel deterministic-latency switch architectures(Xhaul packet Forwarding Element, XFE). Stan-dardization is expected to play a major role in afuture 5G integrated front haul/backhaul architec-ture for multi-vendor interoperability reasons. Tothis end, we review the major relevant activitiesin the current standardization landscape and thepotential impact on the Xhaul architecture.

XHAUL: TOWARD ANINTEGRATED FRONTHAUL/BACKHAUL

ARCHITECTURE IN 5G NETWORKS

DELGADO_LAYOUT.qxp_Author Layout 10/16/15 12:28 PM Page 32

IEEE Wireless Communications • October 2015 33

ing rise to the so-called flexible functional split,where the radio protocol split between central-ized and distributed base station functions canbe adjusted on a case-by-case basis.

In this context, the distinction between front -haul and backhaul transport networks will blur asvarying portions of functionality of 5G points ofattachment (5GPoAs) might be moved towardthe network as required for cost efficiency rea-sons. The traditional capacity overprovisioningapproach on the transport infrastructure will nolonger be possible with 5G. Hence, a new gener-ation of integrated fronthaul and backhaul tech-nologies will be needed to bring capitalexpenditure (CAPEX) and operational expendi-ture (OPEX) to a reasonable return on invest-ment (ROI) range. Also, for cost reasons, theheterogeneity of transport network equipmentmust be tackled by unifying data, control, andmanagement planes across all technologies asmuch as possible. A redesign of the fronthaul/backhaul network segment is a key point for 5Gnetworks since current transport networks cannotcope with the amount of bandwidth required for5G. Next generation radio interfaces, using 100MHz channels and squeezing the bit-per-mega-hertz ratio through massive MIMO or even full-duplex radios, requires a 10-fold increase incapacity on the Xhaul segment, which cannot beachieved just through the evolution of currenttechnologies. Considering the new challengesbrought by 5G, this article presents a novel archi-tecture aimed at integrating fronthaul and back-haul in a common packet-based network definedas Xhaul. This work is relevant due to theincreased interest in architectures integratingboth the fronthaul and backhaul network seg-ments, currently seen as one of the major bottle-necks of the network. The presented architectureis in line with the ideas of recently created stan-dardization bodies, such as the Next GenerationFronthaul Interface Alliance (NGFI) or the workperformed in IEEE 1904.3 and IEEE 802.1. Theremainder of this work is structured as follows.We present the proposed Xhaul architecture withthe key characteristic of integration of fronthauland backhaul network segments into a commonpacket-based framework. Following that, we pre-sent the key challenges that must be addressedfor Xhaul to become a reality. These challengeswill require modifications to key standards andthe development of new ones; hence, we providea brief overview of the expected standardizationimpact. The article concludes with a summary.

XHAUL ARCHITECTURE

To address the aforementioned challenges, we pro-pose the Xhaul architecture, aimed at developingthe next generation of 5G integrated backhaul andfronthaul networks enabling a flexible and soft-ware-defined reconfiguration of all networking ele-ments in a multi-tenant and service-orientedunified management environment. The envisionedXhaul transport network will consist of high-capac-ity switches and heterogeneous transmission links(e.g., fiber or wireless optics, high-capacity copper,or millimeter-wave) interconnecting remote radioheads, 5GPoAs (e.g., macro and small cells),pooled-processing units (mini data centers), and

points of presence (PoPs) of the core networks ofone or multiple service providers.

This requires completely new physical layertechnologies or a radical evolution of existingones, such that the challenging 5G performancerequirements can be met. The Xhaul architecturewill use a novel unified data plane protocol ableto transport both backhaul and fronthaul traffic,regardless of the functional RAN split. As anexample, major challenges for such a protocol arethe big amount of data to handle, the synchro-nization of user data, and reflection of the chan-nel structure of RAN protocols, such as ThirdGeneration Partnership Project (3GPP) LongTerm Evolution (LTE) dedicated traffic channel(DTCH, logical channel), downlink shared chan-nel (DL-SCH, transport channel), and physicaldownlink shared channel (PDSCH, physical chan-nel) [7]. Thereby, the information granularity ofthe RAN protocol should be maintained (e.g.,bearer and user specific information), and com-plemented by the necessary routing information.

Furthermore, the designed unified data planeprotocol must allow for future RAN evolution,which may happen on shorter timescales thantransport network upgrades. The new data planewill also need new switch architectures to sup-port the required latency and jitter demands.Hence, significant effort is needed for the com-bined development of transmission technologies,switch architectures, and protocols in a way thatallows CPRI/OBSAI ([8, 9]) data to be trans-ported in a packetized form over the unifiedXhaul data plane.

It is worth highlighting that full centralizationmight not always be the solution for a particular5GPoA. In particular, outdoor small cells can beextremely difficult to connect to the network, asthey are usually mounted at the street levelinstead of the rooftop level, and many differentobstacles (buildings, trees, etc.) might obstructthe line of sight (LOS) to a transport networkpoint of presence. Hence, in some cases, non-LOS technologies can be the only option to pro-vide small cell connectivity in meshed scenarios.While non-LOS technologies are typically farfrom reaching the capacities required for RANcentralization, they will still be considered forthe Xhaul architecture given their relevancewithin the RAN ecosystem.

Finally, the novel architecture will alsorequire the development of a unified control andmanagement plane (network model and inter-face primitives) to simplify network operationsacross heterogeneous technologies. It is under-stood that the Xhaul architecture must be ableto coexist with legacy infrastructure and mustallow operators smooth migration. The key ideaof the proposed architecture is to move from atraditional transport network perspective to anovel architecture integrating both fronthaul andbackhaul into the Xhaul transport network.

The two key elements of the Xhaul architec-ture are explained in the following and shown inFig. 1.

The Xhaul packet forwarding element (XFE):This element is the central part of the Xhaulinfrastructure, integrating the different physicaltechnologies used for fronthaul and backhaulthrough a common data frame and forwarding

To address these chal-lenges, we propose the

Xhaul architecture,aimed at developing the

next generation of 5Gintegrated backhaul and

fronthaul networksenabling a flexible andsoftware-defined recon-

figuration of all network-ing elements in a

multi-tenant and service-oriented unified manage-

ment environment.


IEEE Wireless Communications • October 201534

behavior. Developing a flexible frame format is akey aspect of fronthaul/backhaul integration,allowing the transport of fronthaul/backhaultraffic on the same physical link, replacing dif-ferent technologies by a uniform transport tech-nology for both network segments.

The Xhaul control infrastructure (XCI): Thiselement is a control infrastructure using a uni-fied abstract network model for control planeintegration. The XCI will be developed byextending existing software defined network(SDN) controllers to provide the services fornovel northbound and southbound interfaces(NBI and SBI), and enable multi-tenancy sup-port in trusted environments. The key aspect ofthe XCI is the development of new mechanismsto abstract the mobile transport network andaggregate measured contextual information.

Figure 1 shows the physical infrastructure ofwhich Xhaul is composed, categorized into threedifferentiated layers. The bottom layer corre-sponds to the interconnection plane and showsthe networking infrastructure, formed by hetero-geneous links connecting the different elementsof the planes located above. Xhaul will typicallyconnect an element of the end-point plane(uppermost plane in Fig. 1) with an element inthe general processing plane (middle plane inFig. 1). This represents the interconnection of,say, a remote radio head implementing radiosampling with a virtualized baseband unit (BBU).

The interconnection plane makes use of Xhaulpacket forwarding elements (XFEs) to intercon-nect a broad set of novel technologies to create apacket-based network that can meet the demandsof 5G networks. The technologies, which havealready been identified as relevant for the future

of backhauling, are represented in the figure too.They span from fiber optics to novel CPRI-over-packet technologies, also considering wirelesslinks such as millimeter-wave (mmWave).

The second plane (depicted in green in Fig. 1)has been named the Xhaul general processingplane and shows the different Xhaul processingunits (XPUs) that carry out the bulk of the opera-tions in Xhaul. The functionality provided bythese XPUs is multi-faceted. It highly depends onthe actual interconnection and must encompassfunctionality expected from the different elementsin the 5G network (depicted in the uppermostlayer of Fig. 1). As an example, in (partially) cen-tralized RAN implementations, the XPUs willhost BBUs or medium access control (MAC) pro-cessors (thus enabling true cloud RAN). Further-more, they might serve as end-points for servicesor even caches. The different functional distribu-tions between 5GPoA and XPU, and the differ-ent services that can be hosted in the XPUs areone of the pillars of the flexibility provided by theXhaul architecture. This is represented by the dif-ferent connection options between the uppermost(end-point plane) and the middle layer of Fig. 1.

Figure 2 shows the Xhaul functional structureand the new Xhaul services. The lowest layercorresponds to the overlay of all infrastructurelayers in Fig. 1. The middle layer represents oneof the key concepts associated with Xhaul: theintegration of the different technologies (includ-ing fronthaul and backhaul) in a common packetnetwork based on technology abstraction, unifiedframing and common data, and the control andmanagement planes. Finally, the upper layerpresents a selection of the features envisionedfor the Xhaul infrastructure:

Figure 1. Xhaul physical infrastructure.

Xhaul packet forwardingelement (XFE)

eNodeBRRHPico/femtoXhaul processing unit (XPU)

Fiber optics

Wireless sub-6 GHzmmWave

MicrowaveCopper

5G PoA

Operator1 PoPOperator2 PoP

CPRI over packet technology

Xhaul interconnection plane

(different infrastructures joined through XFEs)

Fronthaul segment connecting 5GPoA to

XPU

Backhaul segment connecting XPU to

Internet

Xhaul general processing plane

(virtual functions located in XPUse.g., BBUs, CDNs)

Xhaul end-point plane

(e.g., RRH, 5GPoAs, PoPs)

Legacy

DELGADO_LAYOUT.qxp_Author Layout 10/16/15 12:28 PM 4


• Reconfigurability: To cope with the level of demandexpected from 5GPoAs, the Xhaul must be ableto allocate capacity in a dynamic way, reallocatingresources from areas where they are not neededto meet the demands in busier areas (e.g., jointRAN and Xhaul capacity optimization).

• Energy efficiency: This is achieved through tech-niques to reduce the energy consumption of thedifferent Xhaul elements, combined with thejoint optimization of RAN and Xhaul resourcesin terms of dynamic de-activation or decommis-sioning of scarcely used network portions.

• Multi-tenant operation: To enable generalizedsharing and more efficient utilization of theunderlying resources, Xhaul will implementthe concept of multi-tenancy at the infra-structure level. Moreover, Xhaul allows opera-tors and over-the-top (OTT) companies toquickly deploy services within the Xhaul plat-form such as video on demand, HD videocon-ferencing, TV broadcasting, content deliverynetworks (CDNs), and cloud services.The overall Xhaul concept envisioned is

depicted in Fig. 2. Xhaul will integrate the cur-rent functions of backhaul and fronthaul with allthe technologies unified by the use of a commonframing. The data path of the Xhaul will be han-dled by the XFE (on OSI layer 2), interconnect-ing different heterogeneous technologies.

The control and management planes of thenetwork will be handled by a set of controllers(XCIs) interacting with the data plane through anovel SBI capable of supporting the multipletechnologies available in the Xhaul physicallayer. The XCI not only controls packet forward-ing (i.e., the XFE behavior) but also the physicallayer parameters (including optical equipment).The XCI will implement a powerful and openNBI to support a variety of network applicationsrunning dedicated optimization algorithms andfunctions for specific Xhaul features, such ascapacity reconfiguration, forwarding, and infra-structure on demand (IoD) features.

KEY R&D CHALLENGES

This section is devoted to analyzing the key tech-nical challenges to be solved in order to makethe Xhaul architecture a reality.

NOVEL TECHNOLOGIES FORULTRA-HIGH CAPACITY TRANSPORT NETWORKS

In order to cope with the requirements of 5G itis necessary to identify the technologies enablingthe Xhaul network, and understand how thesetechnologies can be combined to realize inte-grated backhaul and fronthaul switching nodes,known as Xhaul forwarding elements (XFEs).Another key aspect is making the differentunderlying technologies transparent to uppercontrol layers according to the SDN paradigm.This poses several outstanding technical chal-lenges. A first challenge is the definition of aunified mechanism for packet and constant bitrate transport, able to deal with the stringentrequirements demanded by the fronthaul links interms of bandwidth and bounded latency. Themechanism should support not only CPRI but beopen to different split options of radio protocolsbetween baseband and remote radio units, mak-ing the challenge even harder. The mechanismwill rely on unified protocol framing for front -haul and backhaul, and the combination of layer1 (L1) and layer 2 (L2) switching for aggrega-tion/disaggregation of multiple fronthaul flowsand Ethernet backhaul traffic over such a uni-versal transport container. A second aspectaddressed is to define how existing technologiesshould be adapted to support the Xhaul conceptand identify new technologies that should beintroduced to fill any gap in terms of latency,bandwidth, or lack of proper infrastructure. Forexample, millimeter waves or wireless opticscould be used when optical fiber or coppercables are not installed. When a fixed access net-work is in place, operators could find that it isconvenient to reuse it for Xhaul purposes, butthis is not trivial because it was typically designedfor much less demanding applications. Finally,greenfield installation opens the door to com-pletely new technologies, such as silicon photon-ics for optical switches, which could cut downthe equipment cost on the order of magnitudes.It is clear from the previous discussion that aone-fits-all approach does not work for Xhaul,but different technologies apply to different usecases. This poses the problem of how to managethese technologies so that the operators are

Figure 2. Xhaul functional structure.

5G PoA

eNodeB

RRH

Pico/femto

Processing node

Fiber optics

Wireless connectionmmWave

MicrowaveGigabit Ethernet

CPRI over novel technologies

n

XHAUL

Reconfigurable

Multi-tenant-enabled

Energy-efficientXhaul services

Technology abstraction and integrationXHAUL data/control and management



unaware of any unnecessary technology details.To accomplish this task it is needed to designand develop the Xhaul SBI, especially develop-ing the protocol and agent extensions to supportthe underlying Xhaul technologies. In the fol-lowing we identify a set of potential technologiesthe development of which will highly impact thecapacity of the Xhaul network segment.

New wireless transmission techniques, such ashybrid RF-optical wireless technologies [10], aimedat data rates as high as 10 Gb/s over 1000 m, astargeted in Xhaul. In outdoor environments wire-less optics is sensitive to haze and fog, whilemmWave faces challenges in rain, so a hybrid solu-tion could be used to increase the link quality. Thewireless optical solutions are robust to electromag-netic interference (EMI) and highly tap-proof,which makes them a good choice in security-awareenvironments. Wireless optics also have the keyadvantage of unlicensed optical spectrum.

Reconfigurable all-optical switches based onsilicon photonics [11]. These devices enhance theflexibility of the Xhaul network with no signifi-cant additional cost, moving to optics the inte-gration concepts well consolidated in electricalintegrated circuits (ICs). This potentially leads totwo orders of magnitude cost reduction withrespect to conventional optical switches.

New flexible multiplexing and aggregationschemes, for example, based on flexi-grid andFlexi-passive optical network (PON). Digital signalprocessing (DSP)-enabled bandwidth-variableoptical transceivers will be considered to interfaceaccess and Xhaul networks, based on multi-carrier(orthogonal frequency-division multiplexing,OFDM) technology for elastic modulation andmultiplexing, introducing finer sub-wavelengthgranularity at the electrical level. This will result infully programmable adaptive optical transceiversthat can achieve optimal bandwidth allocation.

Techniques enabling massive and easydeployment of indoor small cells. More than 80percent of wireless traffic is generated indoors,such as in offices, homes, and shopping malls[12]. Therefore, indoor capacity will be crucialfor 5G success. However, this leads to a numberof technical challenges: high penetration lossinto buildings from the outdoor macro base sta-tions, hostile RF propagation environment, andindoor RF regulations. Current solutions likedistributed antenna systems (DASs) do not meetthe 5G requirements for ultra-high capacity anddensity indoors, so new solutions are needed to:•Pool the baseband resources all over morebuildings• Introduce the capability of dynamic cell splitting

to provide high capacity in hotspot areas andsimultaneous mobility coverage in other areas

• Save energy consumption by switching offunused baseband resources, especially duringclosed hours of these buildings

• New bandwidth-efficient fronthaul solution,with reconfigurability features, to make such adesign easy.

XHAUL UNIFIED DATA PLANE DESIGNThe Xhaul data plane has the ambitious objec-tive of the design of a unified versatile frameformat and the corresponding protocol suite,which supports:

• The traditional split of radio heads and 5GPoAfunctionality

• The concept of moving functionality of the5GPoA among different XPUs

• The concurrent usage of fronthaul traffic (e.g.,OBSAI/ CPRI over fiber) and backhaul traffic(e.g., Ethernet) on the same link

The new frame protocol suite will provide acommon solution for different demands of band-width, latency, and synchronization caused bydifferent split options of the 5GPoA and the dif-ferent technologies available for the fronthaul/backhaul. The main challenge of this task is thecreation of a common data path that can simul-taneously transport backhaul and fronthaul traf-fic with their different characteristics.

Regarding the unified data frame, differentapproaches need to be analyzed to understandthe best way of unifying the different require-ments of the technologies. However, at this time,using a modified version of Ethernet seems to bea promising approach, since multiple productsalready use the Ethernet framing. It is worthhighlighting that traditional Ethernet switchingcannot fulfill the stringent requirements onlatency and jitter for fronthaul traffic [13], sosubstantial optimizations to the switching fabricwill be required for the integration of front -haul/backhaul. Ongoing initiatives (e.g., CPRIover Ethernet, time-sensitive networking [TSN])also need to be analyzed and further developedfor the new frame format to enable a cost-effi-cient and well accepted solution. Alternativeoptions, like the optical transport network(OTN) defined in ITU-T G.709, or OTN-likeframing protocols for the encapsulation andmultiplexing of heterogeneous traffic data willalso be considered. Currently, they present simi-lar problems as Ethernet in dealing with strin-gent latency and jitter requirements.

XHAUL CONTROL PLANE DESIGNAnother of the key challenges faced by theXhaul architecture is the design of a commoncontrol and management plane: the Xhaul con-trol infrastructure (XCI). The XCI will intro-duce an SDN-controlled packet-based networkfor the complete Xhaul network. One of the keyfeatures of the envisioned XCI is to enable thesharing of the Xhaul infrastructure by multipleoperators with differentiated policy and trustedaccess to decrease the total cost of ownership(multi-tenancy). The XCI will also be designedin such a way as to promote the addition of newservices to the transport network in a fast, sim-plified way. The XCI may be implementedthrough an SDN controller and its two maininterfaces, the SBI and NBI. The SBI will pro-vide technology abstraction and adaptation via aset of common control parameters; in this waythe heterogeneous technologies forming theXhaul data plane can be controlled and man-aged in a homogeneous way by using a commonSDN driver to XCI. The XCI will provide anoth-er level of technology abstraction to the applica-tions using the controller through the NBI.

The abstraction at the NBI level allows theSDN controller to deliver mechanisms to dynam-ically program the transport network. The XCIwill collect common contextual information for

Regarding the unifieddata frame, differentapproaches need to beanalysed to understandthe best way of unifyingthe different require-ments of the technolo-gies. However, at thistime, using a modifiedversion of Ethernetseems to be a promisingapproach, since multipleproducts already use theEthernet framing.



traffic load, congestion, interference, and energysavings to retrieve an overall system status of theheterogeneous network. Metrics will be extrapo-lated by new methodologies from a large, versa-tile set of data. The metrics will be utilized bythe Xhaul applications (including performanceoptimization of the network), which will formthe basis of the resource orchestration.

The SBI allows common control and manage-ment of nodes as well as technology- and ven-dor-specific amendments. Several alternatives tothe design of the XCI and the integrationbetween XFE and XCI can be considered, suchas application programming interface (API)calls, plug-ins, or a hierarchical concept of SDNcontrollers. In particular, the hierarchical SDNarchitecture proposed by the Open NetworkFoundation looks very promising as it introducesadditional levels of scalability, modularity, relia-bility, and security, but it must not decrease thelatency or increase the OPEX for the operators.

The key research challenge is to design thecontrol mechanism of the XFEs in such a waythat the different connection demands of front -haul and backhaul can be provided by the samephysical infrastructure. For example, the XCIneeds to reserve sufficient bandwidth for front -haul traffic with no latency and jitter degradationas well as configure the prioritization of the back-haul traffic in the XFEs. Applications of the XCIneed to match the requirements of the differentnetwork elements to the available resources toestablish a working network. Algorithms in XCIanalyze the interaction between fronthaul andbackhaul traffic and lighten the effort with anautomated configuration approach. This matchinghas to follow topology changes due to networkmaintenance as well as dynamic changes of thenetwork triggered by the XCI itself, such as shut-ting down small cells to preserve energy.

XHAUL CONTEXT-AWARERESOURCE ORCHESTRATION

The resources forming the Xhaul network, suchas physical nodes and links, are a major costsource in an operator’s overall infrastructure.With the extreme RAN densification expected tocome along with 5G, total equipment cost isbound to increase even more, putting high pres-sure on operators’ profitability. Furthermore,there are lots of dynamics that need to beaddressed, such as short- and long-term trafficload fluctuations, time-dependent machine-typecommunication patterns, wireless channel varia-tions on the access and transport side, and so on.Therefore, in order to maintain a reasonablecost-revenue balance, it is extremely importantto provide measures that aim at optimallyexploiting the existing network infrastructure atany point of time, avoiding overprovisioning. Inthe Xhaul context, such measures will get a newangle as resource management must be donewith respect to the varying conditions of thewireless access, and the distribution (and redis-tribution) of base station and transport networkfunctions across XPUs. Another factor cominginto play will be energy efficiency (e.g., by shut-ting down base stations, dosing backhaul links,and remapping radio head nodes to BBU pools).

Compared to existing resource managementschemes, Xhaul resource management needsmuch more stringent latency and jitter aware-ness, and much more fine-grained end-to-endquality of service (QoS) differentiation.Indeed, greater diversity of not only 5G userservices, but also transport network services(packet-based phase/frequency/time synchro-nization, CPRI/OBSAI transport, etc.) must beaccommodated. Additionally, the distributionof Xhaul functions not only depends on avail-able node and link resources, but actually canhave a profound impact on the service-levelperformance as well. For example, higher cen-tralization of base station functions results inbetter RAN performance, while more decen-tralization reduces the burden on the transportnetwork.

In summary, the key problem of resourcemanagement in Xhaul is the diversification andhigh number of variables to consider. The RANmay support different levels of centralization(fully or partially centralized, or fully distribut-ed), different backhauling models (in-band, out-of-band, licensed, point-to-point, point-to-multipoint), different physical technologies (lineof sight, copper, fiber) and virtualized parts ofthe RAN and core networks may be located oreven moved to different data centers dependingon their load. This multitude of variables toconsider is bounded by stringent requirementson delay and bandwidth, which affect the rest ofthe components due to the performance interre-lation between RAN and fronthaul/backhaul.Due to this performance interrelation, the gen-eral problem of resource orchestration in Xhaulis the joint optimization of RAN policies (sched-ulers, shapers, handover, etc.), routing policies,and Xhaul function placement (e.g., BBU, virtu-al IP edge functions, mobility anchors, in-net-work processing functions). In the spirit of 5Gambitions, the Xhaul controller must interworkwith controllers of the neighboring networkdomains (i.e., the RAN and core network con-trollers). Ultimately, the exact form of interac-tion will likely be via some global orchestrationengine, with each domain controller (RAN,Xhaul, and core) in turn being controlled bythat engine via a north-south interface.

XHAUL IMPACT ON STANDARDS

The Xhaul architecture’s major enhancementsand new capabilities have the potential for a sig-nificant impact on various standardization bod-ies. In the following we identify threestandardization bodies’ clusters relevant to therealization of the key Xhaul innovations.

Standardization cluster 1 — 3GPP, IEEE802.11, SCF, BBF, NGMN:• New functions and interfaces for both the data

and control planes accounting for variousfunctional splits between the core and RANnetworks

• New underlying wired and wireless technolo-gies that might require amendments to exist-ing or new specifications for the backhaul,fronthaul, or access.Standardization Cluster 2 — ITU-T, Photon-

ics21, FSAN, ETSI, IEEE 1904:

In the spirit of 5G ambi-tions, the Xhaul con-

troller must interworkwith controllers of the

neighboring networkdomains (i.e., the RANand core network con-

trollers). Ultimately, theexact form of interaction

will likely be via some global orchestration

engine, with eachdomain controller in turn

being controlled by that engine via a

north-south interface.



• Novel framing protocols for CPRI opticaltransport or innovative solutions for CPRImapping over Ethernet, OTN, and wave-length-division multiplexing (WDM)

• New enabling technologies for optical front -haulStandardization Cluster 3 — ONF, IETF,

ETSI, IEEE 802.1 OmniRAN, IEEE 802.21:• Novel abstraction of the various underlying

technologies providing the required awarenessat the right level in the Xhaul system architec-ture

• Enhancements on the protocols and architec-tures of SDN and NFV-based systems, takingadvantage of context awareness at variouslocations along the Xhaul path

Within these standardization clusters the rele-vant Xhaul activities per cluster are summarizedin Table 1.

A cohesive and coordinated approach acrossthe aforementioned standardization groups(siloed in nature) would then be desirable tomaximize the Xhaul architecture benefits poten-tial, as shown in Fig. 4.

SUMMARY AND CONCLUSIONS

Next generation RANs will be characterized by ahigher degree of capillarity, density, and higherbit rates. This will push the network segmentsconnecting the RAN to the core of the network,the so-called backhaul, to the limit. At the sametime, new models for the design of base stationsseparating radio and baseband processing in dif-ferent nodes separated geographically and con-nected by specific-purpose communication lines(fronthaul) are being deployed by operators as away of decreasing the OPEX/CAPEX of theirnetworks. With the increase in bandwidthrequired for 5G, non-integrated fronthaul/back-haul architectures are not expected to be able tocope with the new capacity requirements in acost-effective manner.

The Xhaul architecture proposed in this arti-cle is a 5G integrated backhaul and fronthaul

transport network enabling flexible and soft-ware-defined reconfiguration of all networkingelements in a multi-tenant and service-orientedunified management environment. Four keyR&D challenges have been identified to makethe Xhaul architecture a reality:• Novel technologies for ultra-high-capacity trans-

port networks potentially leading to significantadvantages in terms of connectivity, capacityand/or network simplification

• Xhaul unified data plane design of a unified,versatile frame format and the correspondingprotocol suite supporting:–The traditional split of radio heads and5GPoA functionality–The concept of moving functionality of the5GPoA among different XPUs–The concurrent usage of fronthaul traffic andbackhaul traffic on the same link

• Xhaul control plane design of a common con-trol and management plane, the Xhaul controlinfrastructure (XCI), introducing an SDN-controlled packet-based network for the com-plete Xhaul network

• Xhaul context-aware resource orchestration con-sidering the joint optimization of RAN poli-cies, routing policies, and Xhaul functionplacement via a technology-abstracted SDNinterfaceFinally, we have reviewed the relevant stan-

dardization activities that could be impacted bythe Xhaul architecture design and providedinformation regarding the specific groups to betargeted to enable the key Xhaul new functional-ity and features.

ACKNOWLEDGMENTSThe Xhaul architecture is a joint effort of 21companies and academic institutions. Theauthors would like to thank the following per-sons (among many others who also participat-ed) for their contribution to the ideas on thisarticle: Chenguang Lu (Ericsson AB), MiguelAngel Puente (Atos Spain), Luis M. Contreras(Telefonica I+D), Anna Pizzinat (Orange),

Figure 3. Xhaul concept design.

Adaptation layer(common packetization)

LTE

mmWave

CPRI

IP-based traffic

Xhaul forwarding element

Xhaul data path

Xhaul forwarding element

Xhaul data pathXhaul data path

Operator core

Novel RAN andtransport technologies

Xhaulcontrol logic

Xhaul processing units

Services over Xhaul (e.g., CDN)

5GPoA functionality

Local breakout

Applications:- Capacity reconfiguration- Energy optimization

XH

AU

L D

ata

Path

Xhaul management and control plane

Next generation radioaccess networks will becharacterized by a higher degree of capillarity, density andhigher bit rates. This willpush to the limit thenetwork segments connecting the RAN tothe Core of the network,the so called backhaul.



David Jimenez (Visiona IP), Laurent Bellot (e-Blink), Giada Landi (NextWorks), Dragos Vin-garzan (Core Network Dynamics), Jose V.Galan (TELNET), Josep Mangues-Bafalluy(CTTC), Tinku Rasheed (CREATE-NET),Claudio Casetti (Politecnico di Torino), PerOdling (Lunds University), and Fang-Chu Chen(ITRI). This work has been partly supported bythe Madrid Regional Government through theTIGRE5-CM program (S2013/ICE-2919) andthe EU H2020 Xhaul Project (grant no.671598).

REFERENCES[1] Cisco, “Cisco Visual Networking Index: Forecast and

Methodology, 2013–2018,” white paper, June 2014;http://www.cisco.com/c/en/us/solutions/collateral/ser-v i c e - p r o v i d e r / i p - n g n - i p - n e x t - g e n e r a t i o n -network/white_paper_c11-481360.pdf.

[2] D. Astely et al., “LTE Release 12 and Beyond,” IEEECommun. Mag., vol. 51, no. 7, 2013, pp. 154–60.

[3] A. Pizzinat et al., “Things You Should Know aboutFront haul,” IEEE/OSA J. Lightwave Tech., 2015.

[4] “C-RAN: The Road towards Green RAN,” China Mobilewhite paper, v. 2, 2011.

[5] M. Nahas et al., “Base Stations Evolution: Toward 4GTechnology,” 2012 19th IEEE Int’l. Conf. Telecommun.,2012, pp. 1–6.

Figure 4. Xhaul architecture impact on standardization bodies.

New functions and interfacesfor both the data and controlplanes accounting for variousfunctional splits between thecore and the RAN networks

Novel framing protocol for CPRIoptical transport or innovativesolutions for CPRI mapping overEthernet, OTN and WDM

Novel abstraction of the variousunderlying technologiesproviding the right level ofawareness at the right level inthe Xhaul system architecture

Enhancements on the protocolsand architectures of SDN andNFV-based systems, takingadvantage of contextawareness at various locationsalong the Xhaul

New enabling technologies forthe fronthaul optical interface(both fixed and mobile)

New underlying wirelesstechnologies that might requireamendments to existing or newspecifications for the backhaul,fronthaul or access

IEEE802.1cf

FSANITU-TSG15

IEEE802

ETSIORI

NGMN

BBFSCF

XHAUL relevant standardization bodies

IEEE802.11

3GPP

Photonics21

ETSIMEC

ETSINFV

IRTFONF IETF

Table 1. Standardization clusters and corresponding relevant activities.

Standardization Cluster 1 Activities

Third Generation Partnership Project (3GPP)IEEE 802.11Small Cell Forum (SCF)Broadband Forum (BBF)Next Generation Mobile Networks (NGMN)

SA1; SA2; SA5; RAN1; RAN2; RAN4.HEW; NG60Backhaul; Network; Radio & Physical Layer; Services; DeploymentSIM-TD; E2E Architecture and Service InnovationSmall Cells; RAN evolution; ORI; 5G initiative


International Telecommunications Union —Telecommunications Standardization Sector (ITU-T)Photonics21Full Service Access Network (FSAN)European TelecommunicationsStandards Institute (ETSI)IEEE 1904

SG15

WG1 Information and CommunicationNext Generation PON (NG-PON) Task GroupOpen Radio Interface (ORI) ISG

1904 RoE


Open Networking Foundation (ONF)

Internet Engineering Task Force (IETF)/Internet ResearchTask Force (IRTF)ETSI

IEEE 802

WMWG; Mobile & Wireless; Optical Transport; Carrier Grade SDN; Migra-tion; L4-L7; North Bound InterfacesDMM; I2RS; ForCES; NETCONF; SFC; PCE; ACTN; VNFPool; SDNRG; NVO3;NETEXT; INTAREANetwork Function Virtualization (NFV) ISG; Mobile Edge Computing (MEC)ISG802.1CF; 802.21



[6] P. Rost et al., “Cloud Technologies for Flexible 5G RadioAccess Networks,” IEEE Commun. Mag., vol. 52, no. 5,May 2014, pp. 68–76.

[7] 3GPP, “Technical Specification Group Radio Access Net-work; Evolved Universal Terrestrial Radio Access (E-UTRA); Medium Access Control (MAC) ProtocolSpecification, Release 11,” Oct. 2012; http://www.qtc.jp/3GPP/Specs/36321-b10.pdf.

[8] “V6.1 Common Public Radio Interface (CPRI); InterfaceSpecification,” July 2014; Ericsson AB, Huawei Tech-nologies Col Ltd, NEC Corporation, Nortel Networks SAand Siemens Networks Gmbh & Co,” Ltd, NEC Corpora-tion, Nortel Networks SA, Siemens Networks GmbH &Co. KG, Ericsson AB, and Huawei Technologies Co Ltd.

[9] “OBSAI System Specification.”[10] V. Jungnickel, “High-Speed Optical Wireless Communi-

cations Technologies,” OFC ’14, 2014, pp. Th1F–5.[11] A.-J. Lim et al., “Review of Silicon Photonics Foundry

Efforts,” IEEE J. Sel. Topics Quantum Electron., vol. 20,no. 4, 2014, pp. 405–16.

[12] “Ericsson Mobility Report,” June 2014; http://www.ericsson.com/res/docs/2014/ericsson-mobility-report-june-2014.pdf.

[13] J. Aweya, “Implementing Synchronous Ethernet inTelecommunication Systems,” IEEE Commun. Surveys &Tutorials, vol. 16, no. 2, 2nd qtr. 2014, pp. 1080–113.

BIOGRAPHIESANTONIO DE LA OLIVA ([email protected]) received histelecommunications engineering degree in 2004 and hisPh.D. in 2008 from UC3M, where he has been working asan associate professor since then. His current line ofresearch is related to networking in extremely dense net-works. He is an active contributor to IEEE 802 where hehas served as Vice-Chair of IEEE 802.21b and Technical Edi-tor of IEEE 802.21d. He has also served as a Guest Editorof IEEE Communications Magazine. He has published morethan 30 papers on different areas of networking includingmobility and wireless networks.

XAVIER COSTA PÉRÉZ is the technical manager of the Xhaulproject and manager of the 5G Networks R&D Group atNEC Laboratories Europe. At NEC he manages several pro-jects focused on 5G mobile core, backhaul/fronthaul, andaccess networks, both internally for products roadmap evo-lution and externally in EU collaborative projects. His teamhas received several R&D awards for successful producttransfers and holds over 20 patents. He received both hisM.Sc. and Ph.D. degrees in telecommunications from thePolytechnic University of Catalonia.

ARTURO AZCORRA received his M.Sc. degree in telecommuni-cations engineering from the Universidad Politécnica deMadrid (UPM) in 1986 and his Ph.D. from the same univer-sity in 1989. In 1993, he obtained an M.B.A. with honorsfrom Instituto de Empresa. He has participated in anddirected 49 research and technological development pro-jects, including the European ESPRIT, RACE, ACTS, and ISTprograms. He has coordinated the CONTENT and E-NEXTEuropean Networks of Excellence, and the CARMEN andXhaul EU projects. He has served as a Program Committeemember for many international conferences, including sev-eral editions of IEEE PROMS, IDMS, QofIS, ACM CoNEXT,and IEEE INFOCOM. He is the founder of the ACM CoNEXTconference series and was the General Chair of its first edi-tion. He has published over 100 scientific papers in books,international magazines, and conferences. In addition tohis scientific achievements, he has a relevant track recordof research management. He was deputy vice rector ofacademic infrastructures at University Carlos III of Madrid(UC3M) from 2000 to 2007. He served as Director Generalfor Technology Transfer and Corporate Development at theSpanish Ministry of Science and Innovation from 2009 to2010, and was then appointed Director General of CDTI(the Spanish agency for industrial research) from 2010 to2012. He is the founder of the international research cen-ter IMDEA Networks, and currently is its director, with adouble affiliation as a full professor at UC3M.

ANDREA DI GIGLIO received a Dr. Ing. degree in electronicengineering from the University of Pisa, Italy, and ScuolaSanta Anna. He is with Telecom Italia Lab. His fields ofresearch are Internet security and optical networks archi-tecture. He has been involved in several European Projects:the architectural Work Package of the IST Project NOBEL,DISCUS as Work Package leader, and ICT STRONGEST asproject coordinator.

FABIO CAVALIERE is with Ericsson, Sweden, where he has beensince 2006. As an expert in photonic systems and technolo-gies, he contributes to developing strategy in optical trans-port and access networks. Before joining Ericsson, he waswith Marconi, United Kingdom, from 1998 to 2006 in thePhotonic Network and System Design Authority. He is co-editor of ITU-T Recommendation G.metro and a member ofthe Board of Stakeholders of Photonics 21. He has authoredseveral publications and patents on optical communications.

DIRK TIEGELBEKKERS is a senior system specialist in the trans-port system architecture group for LTE. He has around 18years of experience in transport network protocols formobile wireless communication systems, mainly IP-basedbackhaul networks. He has actively contributed in 2005 tothe Nokia specification team for 3.9G by analyzing the per-formance of MAC/ RLC proposals. In 2008, he was highlyinvolved in the specification of Nokia’s Packet Abis solu-tion, which enables operators to employ packet switchednetworks as a means to provide the transport backhaul fornetwork traffic. He served as technical manager and pro-ject lead for several R&D projects for Nokia Research Centreand Nokia Networks. His current focus area is LTE-A and itsimpact on mobile transport.

JOHANNES LESSMANN is a senior researcher in the Mobile andWireless Networks group of NEC Laboratories Europe, Hei-delberg, Germany. He currently serves as project managerfor internal as well as customer and EU R&D projects in thedomain of mobile backhaul and core networks. His majorresearch topics include network planning, end-to-endresource and QoS management, SDN for transport networks,and mobile carrier cloud. He has authored many standardscontributions, and journal and conference publications.

THOMAS HAUSTEIN received his Dr.-Ing. (Ph.D.) degree inmobile communications from the University of TechnologyBerlin, Germany, in 2006. In 1997, he was with the Fraun-hofer Institute for Telecommunications, Heinrich HertzInstitute (HHI), Berlin, where he worked on wirelessinfrared systems and radio communications with multipleantennas and OFDM. He focused on real-time algorithmsfor baseband processing and advanced multiuser resourceallocation. From 2006 to 2008, he was with Nokia SiemensNetworks, where he conducted research for 4G. Since 2009he has headed the Wireless Communications Departmentat Fraunhofer HHI focussing on 5G and Industrial Wireless.

ALAIN MOURAD is currently leading the R&D of 5G wirelesssystems at InterDigital Europe Ltd, United Kingdom. Priorto joining InterDigital, he was a principal engineer at Sam-sung Electronics R&D, United Kingdom, and previously asenior engineer at Mitsubishi Electric R&D Centre Europe,France. Throughout his career, he has been active in theresearch and standardization of recent wireless communi-cation and broadcasting systems (3GPP LTE, WiMAX 16m,DVB-T2/NGH, ATSC 3.0). He has several inventions and hasauthored numerous peer-reviewed scientific publications.He received the Inventor of the Year Award from SamsungElectronics R&D UK in 2012 and 2013.

PAOLA IOVANNA received her degree in electronics engineeringfrom the University of Roma “Tor Vergata” in 1996. From1995 to 1997 she collaborated with the FUB research centerof Rome, working on fiber optic communications and opti-cal networking. From 1997 to 2000 she worked at TelecomItalia, where she was involved in experimentation of newservices based on different access technologies (e.g., XDSL,frame relay, optical). In 2000 she joined Ericsson in theResearch Department where she dealt with networking anddesign solutions for packet and optical technology ( GMPLS,MPLS, Ethernet ). From 2009 to 2012 she was responsiblefor carrying out research projects on packet and opticalrouting, control plane, and path computation solutions. Inthe framework of such activities she realized demonstratorsand prototypes as well. In 2012 she was responsible fordefining and prototyping SDN solutions for multi-domaintransport in collaboration with customers. Since 2014 shehas lead a research team to define transport networking andcontrol solutions for 5G. She is actively involved in Europeanprojects and is on the Technical Program Committees ofinternational conferences like ECOC. She holds more than 40patents in routing, traffic engineering systems, and PCE solu-tions for packet-opto networks based on GMPLS, and multi-domain SDN transport, fronthaul, and backhaul solutions for5G, and is an author of several tens of publications in eitherinternational scientific journals or conferences.

The Xhaul architectureproposed in this articleis a 5G integrated backhaul and fronthaultransport networkenabling a flexible andsoftware-defined reconfiguration of allnetworking elements ina multi-tenant and service oriented unifiedmanagement environment.


smart backhauling and fronthauling 5g networks …

Documents