grupo 4 desizxcgn considerations for a 5g network architecture

IEEE Communications Magazine November 2014 650163-6804/14/$25.00 2014 IEEE

Patrick KwadwoAgyapong, Mikio Iwamu-ra, Dirk Staehle, andWolfgang Kiess are withDOCOMO Communica-tions Laboratories EuropeGmbH.

Anass Benjebbour is withNTT DOCOMO Inc.

1 In [3], reliability isdefined as the probabil-ity that a certain amountof data to or from an enduser device is successful-ly transmitted to anotherpeer (e.g., Internet serv-er, mobile device, sensor,etc.) within a predefinedtime frame, i.e., before acertain deadline expires.The amount of data tobe transmitted and thedeadline are dependenton the service character-istics.

2 In the context of thisarticle, robustness isdefined as the ability ofthe network to support aminimum predefinedservice level (e.g., mini-mum signal-to-interfer-ence-plus-noise ratio,SINR, to support basicvoice communications)regardless of the networkconditions (e.g., in natu-ral disasters).

INTRODUCTIONDespite the advances made in the design andevolution of fourth generation cellular networks,new requirements imposed by emerging commu-nication needs necessitate a fifth generation(5G) mobile network. New use cases such ashigh-resolution video streaming, tactile Internet,road safety, remote monitoring, and real-timecontrol place new requirements related tothroughput, end-to-end (E2E) latency,reliability,1 and robustness2 on the network. Inaddition, services are envisioned to provideintermittent or always-on hyper connectivity formachine-type communications (MTC), coveringdiverse services such as connected cars, connect-ed homes, moving robots, and sensors that mustbe supported in an efficient and scalable man-ner. Furthermore, several emerging trends suchas wearable devices, full immersive experience(3D), and augmented reality are influencing thebehavior of human end users and directly affect-ing the requirements placed on the network. Atthe same time, ultra-dense small cell deploy-ments and new technologies such as massivemultiple-input multiple-output (mMIMO), soft-ware defined networking (SDN), and networkfunction virtualization (NFV) provide an impe-

tus to rethink the fundamental design principlestoward 5G.

This article proposes a novel 5G mobile net-work architecture that accommodates the evolu-tion of communication types, end-user behavior,and technology. The article first highlights trendsin end-user behavior and technology to motivatethe challenges of 5G networks. Some potentialenablers are identified, and design principles fora 5G network are highlighted. This is followedby the articulation of a 5G mobile network archi-tecture together with details of some fundamen-tal technology enablers and design choices, anda discussion of issues that must be addressed torealize the proposed architecture and an overall5G network. The article wraps up with proof ofconcept evaluations and conclusions.

CURRENT TRENDSIt is well known that mobile data consumption isexploding, driven by increased penetration ofsmart devices (smartphones and tablets), betterhardware (e.g., better screens), better user inter-face design, compelling services (e.g., videostreaming), and the desire for anywhere, anytimehigh-speed connectivity. What is perhaps notwidely mentioned is that more than 70 percentof this data consumption occurs indoors inhomes, offices, malls, train stations, and otherpublic places [1]. Furthermore, even thoughmobile data traffic is increasing at a brisk pace,signaling traffic is increasing 50 percent fasterthan data traffic [2].

More end users are using multiple deviceswith different capabilities to access a mix of besteffort services (e.g., instant messaging and email)and services with quality of experience (QoE)expectations (e.g., voice and video streaming).Over-the-top (OTT) players provide services andapps, some of which compete directly with coreoperator services (e.g., voice, SMS, and MMS).Connectivity is increasingly evaluated by endusers in terms of how well their apps work asexpected, regardless of time or location (in acrowd or on a highway), and they tend to beunforgiving toward the mobile operator whenthese expectations are not met. Moreover, thebattery life of devices and a seamless experienceacross multiple devices (or a device ecosystem)

ABSTRACTThis article presents an architecture vision to

address the challenges placed on 5G mobile net-works. A two-layer architecture is proposed, con-sisting of a radio network and a network cloud,integrating various enablers such as small cells,massive MIMO, control/user plane split, NFV,and SDN. Three main concepts are integrated:ultra-dense small cell deployments on licensedand unlicensed spectrum, under control/userplane split architecture, to address capacity anddata rate challenges; NFV and SDN to provideflexible network deployment and operation; andintelligent use of network data to facilitate opti-mal use of network resources for QoE provision-ing and planning. An initial proof of conceptevaluation is presented to demonstrate thepotential of the proposal. Finally, other issuesthat must be addressed to realize a complete 5Garchitecture vision are discussed.

5G NETWORKS: END-TO-END ARCHITECTURESAND INFRASTRUCTURE

Patrick Kwadwo Agyapong, Mikio Iwamura, Dirk Staehle, Wolfgang Kiess, and Anass Benjebbour

Design Considerations for a 5G Network Architecture

AGYAPONG_LAYOUT.qxp_Layout 10/29/14 3:33 PM Page 65

IEEE Communications Magazine November 201466

have also become important issues for many endusers.

The Internet of Things (IoT), which addsanything as an additional dimension to connec-tivity (in addition to anywhere and anytime), isalso becoming a reality. Smart wearable devices(e.g., bracelets, watches, glasses), smart homeappliances (e.g., televisions, fridges, thermostats),sensors, autonomous cars, and cognitive mobileobjects (e.g., robots, drones) promise a hyper-connected smart world that could usher in manyinteresting opportunities in many sectors of lifesuch as healthcare, agriculture, transportation,manufacturing, logistics, safety, education, andmany more. Even though operators currently relyon existing networks (especially widely deployed2G/3G networks and fixed line networks) to sup-port current IoT needs, many of the envisagedapplications impose requirements, such as, verylow latency and high reliability, that are not easi-ly supported by current networks.

To cope with such evolving demands, opera-tors are continuously investing to enhance net-work capability and optimize its usage. Operatorsare deploying more localized capacity, in theform of small cells (e.g., pico and femto cellsand remote radio units, RRUs, that are connect-ed to centralized baseband units by optical fiber)to improve capacity. In addition, traffic offload-ing to fixed networks through local area tech-nologies such as Wi-Fi in unlicensed frequencybands has become widespread. To optimize net-work usage for better QoE in a fair manner,mobile networks are also integrating more func-tionality such as deep packet inspection (DPI),

caching, and transcoding. All these improve-ments come at significant capital and operatingcosts, however.

With the increasing complexity and associatedcosts, several concepts and technologies that haveproved useful to the information technology (IT)sector are becoming relevant to cellular networksas well. For instance, an industry specificationgroup (ISG) set up under the auspices of the Euro-pean Telecommunications Standards Institute(ETSI ISG NFV) is currently working to definethe requirements and architecture for the virtual-ization of network functions and address identifiedtechnical challenges. Similarly, the Open Network-ing Foundation approved a Wireless and MobileWorking Group in November 2013 to identify usecases in the wireless and mobile domain that canbenefit from SDN based on OpenFlow.

5G CHALLENGES, ENABLERS, ANDDESIGN PRINCIPLES

Based on current trends, it is generally under-stood that 5G mobile networks must address sixchallenges that are not adequately addressed bystate-of-the-art deployed networks (Long TermEvolution-Advanced, LTE-A): higher capacity,higher data rate, lower E2E latency, massivedevice connectivity, reduced capital and opera-tions cost, and consistent QoE provisioning [3,4]. These challenges are briefly discussed belowtogether with some potential enablers to addressthem. Figure 1 provides an overview of the chal-lenges, enablers,3 and corresponding design prin-

Figure 1. 5G challenges, potential enablers, and design principles.

Capacity

x1000> 70% indoor

Data rate

x10-100

E2E latency

< 5ms

Cost

Sustainable

QoE

Consistent

Massivenumber of

connections

x10-100

Spectrum

5G design principlesEnablers to address challengesChallenges

Use high frequencies and other spectrumoptions (e.g., pooling, aggregation).

C/U-plane split Address coverage and capacity separately.

NFV/SDN/cloud Minimize number of network layers andpool resources as much as possible.

3rd party/userdeployment models

Simple access points

Minimize functionalities performed byaccess points.

Energy-efficienthardware

Energy managementtechniques

Maximize energy efficiency across allnetwork entities.

All optical networks Optical transmission and switchingwherever possible.

Small cells

Local offload (e.g., D2D, enhanced local area)

Caching/pre-fetching/CDN

Bring communicating endpoints closertogether.

SON

Traffic management

Big data-drivennetwork intelligence

Use an intelligent agent to manage QoE,routing, mobility and resource allocation.Redesign NAS protocols, services andservice complexity.

Massive/3D MIMO

New air interface(e.g., new waveform, advanced

multiple access, shorter TTI)

Design new air interface, new multipleaccess scheme and L1/L2 techniques thatcan be optimized for high frequencies,latency and massive connectivity.

Based on current

trends, it is generally

understood that

5G mobile networks

must address six

challenges that are

not adequately

addressed by state-

of-the-art deployed

networks: higher

capacity, higher data

rate, lower E2E

latency, massive

device connectivity,

reduced capital and

operations cost,

and consistent

QoE provisioning.

3 The connectionsbetween the challengesand enablers depict themost significant linking,but not necessarily allpossible connections.


IEEE Communications Magazine November 2014 67

ciples for 5G. It must be noted that the enablershighlighted in Fig. 1 also introduce their own setof challenges and corresponding key perfor-mance indicators (KPIs). Some of these chal-lenges are discussed in the relevant sections.Nevertheless, a detailed discussion of the rele-vant KPIs is outside the scope of this article.The interested reader is referred to [4] for moredetails on this aspect.

SYSTEM CAPACITY AND DATA RATEBeyond 2020 mobile networks need to support a1000-fold increase in traffic relative to 2010 lev-els, and a 10- to 100-fold increase in data rateseven at high mobility and in crowded areas ifcurrent trends continue [1, 3, 4]. This requiresnot only more capacity in the radio access net-work (RAN), but equally important, also in thebackbone, backhaul, and fronthaul. Pricingschemes can be used to manage and potentiallyreduce the increase in data consumption, asalready demonstrated by operators in the mar-ket. However, as customers are willing to pay forthe provisioned service rather than the data vol-ume, pricing models may not be effective to sup-press traffic in the future.

The current consensus is that a combination ofmore spectrum, higher spectrum efficiency, net-work densification, and offloading are necessary toaddress these challenges in the RAN [5]. Opportu-nities for more spectra include higher frequencybands (e.g., millimeter-wave, mmW), unlicensedspectrum, and aggregation of fragmented spectrumresources using carrier aggregation techniques.Dual connectivity of terminals to multiple base sta-tions can exploit aggregated use of spectrumdeployed at different base stations. Besides theavailable bandwidth, high frequency bands alsoallow for mMIMO using antenna arrays with smallform factors, which can provide a 10-fold increasein capacity compared to conventional single-anten-na systems [6]. Nevertheless, high frequency bandssuffer from high path loss attenuation and are lim-ited to line of sight (LOS) and short-range non-LOS environments. Massive MIMO can beexploited to extend the coverage of higher frequen-cy bands by relying on beamforming gains.

Advanced physical layer techniques, such ashigher-order modulation and coding schemes(MCS), such as 256-quadrature amplitude mod-ulation (QAM), increase spectral efficiency andcan be combined with mMIMO to increase sys-tem capacity. By adding some intelligence at thetransmitter and receiver, potential interferencecan be coordinated and cancelled at the receiverto increase system throughput [7]. With suchtechniques in place, new schemes such as non-orthogonal multiple access (NOMA), filter bankmulticarrier (FBMC), and sparse coded multipleaccess (SCMA) can further be utilized toimprove spectral efficiency. For example, NOMAwith successive interference cancelling (SIC)receivers has been shown to improve overallthroughput in macrocells compared to orthogo-nal multiple access schemes by up to 30 percenteven for high-speed terminals, with further gainsexpected with advanced power control [8].

Network densification refers to the densedeployment of many small cells. High carrier fre-quencies are well suited for small cells. The high

attenuation they suffer is no longer seen as adrawback, but rather as an enabler to provideeffective separation and mitigate interferencebetween densely deployed small cells. To allowefficient improvement of capacity at critical loca-tions, it is desirable that coverage and capacitybe addressed independently. This can be realizedthrough an architecture where control (C) anduser data (U) planes are split among differentcells [9]. The benefit of this approach is that U-plane resources can be scaled independent of C-plane resources. This allows more U-planecapacity to be provided in critical areas where itis needed, without the need to also provide co-located C-plane functionalities. Thus, more flexi-ble deployments at lower costs can be realized.In such a C/U-plane split architecture, macro-cells can provide coverage (C+U), and smallcells can provide localized capacity (U).

Techniques like mMIMO and higher-orderMCS can be employed in small cells to boostthroughput [5]. Massive MIMO has an increasedrisk of link failure due to narrow beamforming,but this could be mitigated by employing robusttechniques like dual connectivity, which alwaysprovides uninterrupted fallback to the coveragelayer. Additionally, local offload through tech-niques such as network-controlled device-to-device (D2D) communications can furtherincrease achievable system throughput [10].

Advances in optical networking, includingoptical switching, may be able to address thecapacity requirements in the backbone, back-haul, and fronthaul. In addition, mMIMO can beused to provide high-capacity wireless backhauland fronthaul links in LOS conditions.

END-TO-END LATENCYEnd-to-end latency is critical to enable new real-time applications. For example, remote con-trolled robots for medical, first response, andindustrial applications require rapid feedbackcontrol cycles in order to function well. Safety-critical applications for cars and humans, builtaround vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communication, also requirevery quick request-response and feedback con-trol cycles with high availability and reliability.Augmented and virtual reality applications (e.g.,immersive displays and environments) requirevery fast request-response cycles to mitigatecyber sickness. In order to realize these applica-tions, networks must be able to support a targetof 1 ms E2E latency with high reliability [11].

Innovations in air interface, hardware, protocolstack, backbone, and backhaul (all-optical trans-mission and switching), as well as network archi-tecture can all help to meet this challenge. A newair interface with new numerology, such as shortertransmission time interval (TTI), can reduce over-the-air latency to a few hundred microseconds.Shorter TTI requires high available bandwidth,but this can be supported by using higher frequen-cy bands. Note that such new numerology relieson significant improvements in receiver hardware(e.g., processing power and buffer size).

In addition, E2E latency can be reduced byenhancements in higher-layer protocols (e.g., usecase and network-aware admission/congestioncontrol algorithms to replace TCP slow start),

To allow efficient

improvement of

capacity at critical

locations, it is desir-

able that coverage

and capacity be

addressed indepen-

dently. This can be

realized through an

architecture where

control (C) and user

data (U) planes are

split among

different cells.



bringing communicating endpoints closer (e.g.,through network-controlled D2D and ultra-densesmall cell deployments with local breakout) andadding more intelligence at the edge of the net-work. The latter is realized, e.g., through cachingand pre-fetching techniques, service-dependentlocation of C-plane protocols and orchestration.For example, C-plane protocols necessary forlatency-critical MTC services may be distributedat the edge of the network, whereas C-plane pro-tocols required for services with more relaxedlatency requirements could be located at a cen-tral entity. Efficient design of the non-access stra-tum (NAS) could also help reduce E2E latency.For example, integrating NAS and access stratum(AS) could reduce the control signaling requiredto set up and maintain a data connection, whichcan reduce the E2E latency. Alternatively, devel-oping NAS protocols better tailored to new usecases could also yield a similar result.

MASSIVE NUMBER OF CONNECTIONSThe number of connected devices is expected toincrease between 10- and 100-fold beyond 2020[3]. These will range from devices with limitedresources that require only intermittent connec-tivity for reporting (e.g., sensors) to devices thatrequire always-on connectivity for monitoringand/or tracking (e.g., security cameras, transportfleet). In addition to the sheer number of con-nected devices, a challenge is to support thediversity of devices and service requirements in ascalable and efficient manner.

A combination of advances in air interfacedesign, signaling optimization, and intelligentclustering and relaying techniques can all con-tribute to support hyperconnectivity. Forinstance, using one device as a gateway or relayto aggregate traffic from multiple devices canreduce the signaling load on the network. Moreefficient protocols that combine AS and NASalso reduce the signaling burden. Moreover, con-tention-based and connectionless access proce-dures can be used to efficiently support MTCapplications that only require intermittent con-nectivity to transmit small packets.

Not all devices may be equipped with high-pre-cision devices to cope, for example, with tight syn-chronization to maintain orthogonality of signalsin a multiple access environment when newnumerology is introduced to reduce latency. Tomitigate this, new waveforms such as FBMC,which can suppress out-of-band emission to reduceinterference under an asynchronous environment,can be explored [12]. FBMC also has a potentialto cope better than OFDM with doubly dispersivechannels when both the transmitting and receivingendpoints are moving (e.g., in a V2V application).

In addition, supporting devices with limitedresources such as sensors will require advances inbattery and energy harvesting technologies onone hand and efficient signaling and data trans-mission protocols on the other. For instance,robust medium access techniques combining bothcontrol and data transmission could be explored.

COSTConnectivity is seen as an important enabler forsocio-economic development. Therefore, it isimportant to reduce the infrastructure cost as

well as the costs associated with their deploy-ment, maintenance, management, and operationto make connectivity a universally available,affordable, and sustainable utility. The challengefor the design of 5G is that huge improvementsare needed to address the new requirements, butcustomers are not willing to pay proportionally.In effect, 5G should be a network (RAN, core,backbone routers, and backhaul) that addressesall the new requirements at a cost that will makeservice provisioning sustainable.

Solving the capacity and data rate challengeswith network densification could be very expen-sive in terms of equipment, maintenance, andoperations. One way to reduce equipment cost isto minimize the number of functionalities at thebase station. This could be done by implementingonly layer 1/2 (L1/L2) functionalities in the basestation and moving higher-layer functionalities toa network cloud that serves many base stations.Reducing the number of functionalities results insimpler base stations, which could be deployed byusers and remotely or autonomously managed toreduce deployment and operation costs.

Energy consumption is a significant opera-tions cost driver, with the RAN estimated toconsume 7080 percent of the energy require-ments [13]. Therefore, intelligent energy man-agement techniques, especially in the RAN,could provide a viable means to reduce overallnetwork operations costs. Energy-efficient hard-ware design, low-power backhaul, and intelligentenergy management techniques, especially inultra-dense networks, to put base stations tosleep when not in use can all contribute to reduc-ing the cost of operating a 5G network [13].

NFV and SDN are also viable enablers toreduce costs. NFV decouples network function-ality from dedicated hardware and promotesimplementation of functionality in software ongeneral-purpose IT hardware operated accordingto a cloud model [14]. SDN decouples C- and U-planes of network devices, and provides a logi-cally centralized network view and control, whichfacilitates transport network optimization. Thesetechnologies will make the network more flexibleas new functionality can be introduced with sim-ple software upgrades, and more sophisticatedalgorithms can be employed to manage the net-work from a holistic viewpoint. Moreover,pooled hardware resources can be shared amongmultiple functions, thus realizing multiplexinggains and lowering the amount of necessaryhardware. The flexibilities enabled by NFV andSDN can make the network quick to deploy andmore adaptable, and reduce time to market fornew services.

QOEQuality of experience describes the subjectiveperception of the user as to how well an applica-tion or service is working. Quality of experienceis highly application- and user-specific, and can-not be generalized. For example, the QoE ofvideo applications depends on the quality of theencoded and delivered video in the context ofthe display on which the video is shown. Deliver-ing an application with too low QoE leads touser dissatisfaction, whereas too high QoEunnecessarily drains resources on both the user

Pooled hardware

resources can be

shared among multi-

ple functions, thus

realizing multiplexing

gains and lowering

the amount of nec-

essary hardware. The

flexibilities enabled

by NFV and SDN can

make the network

quick to deploy and

more adaptable, and

reduce time to mar-

ket for new services.



(e.g., device battery) and operator (e.g., radioand transport network resource, base stationpower) sides. Hence, a challenge for 5G is tosupport applications and services with an opti-mal and consistent level of QoE anywhere andanytime.

Despite the diversity of QoE requirements,providing low latency and high bandwidth gener-ally improves QoE. As such, most enablers men-tioned previously can improve QoE. Additionally,traffic optimization techniques can be used tomeet increasing QoE expectations. Furthermore,installing caches and computing resources at theedge of the network allows an operator to placecontent and services close to the end user. Thiscan enable very low latency and high QoE fordelay-critical interactive services such as videoediting and augmented reality.

Better models that describe the relationshipof QoE to measurable network service parame-ters (e.g., bandwidth, delay) and context parame-ters (e.g., device, user, and environment) arealso emerging. Big data, including informationfrom sensors (e.g., on the device) and statisticaluser data, can be used intelligently with suchmodels to more precisely assess the QoE expect-ed by a user and determine the optimal resourcesto use to meet the expected QoE. SDN can thenbe used to flexibly provision the necessaryresources.

Besides the mobile network, advances in thefixed network and potential convergence of thefixed and mobile networks are also needed to

address the challenges highlighted above. How-ever, specific discussions related to the fixed net-work and convergence of the mobile and fixednetworks are outside the scope of this article.

5G MOBILE NETWORKARCHITECTURE VISION

Figure 2 illustrates a 5G mobile network archi-tecture that utilizes the enablers discussed previ-ously. The key elements in the architecture aresummarized below: Two logical network layers, a radio network

(RN) that provides only a minimum set ofL1/L2 functionalities and a network cloudthat provides all higher layer functionalities

Dynamic deployment and scaling of func-tions in the network cloud through SDNand NFV

A lean protocol stack achieved throughelimination of redundant functionalitiesand integration of AS and NAS

Separate provisioning of coverage andcapacity in the RN by use of C/U-planesplit architecture and different frequencybands for coverage and capacity

Relaying and nesting (connecting deviceswith limited resources non-transparently tothe network through one or more devicesthat have more resources) to support multi-ple devices, group mobility, and nomadichotspots

Figure 2. 5G mobile network vision and potential technology enablers.

Lean protocol stack

OTT

Gateway U-plane mobility anchor OTA security provisioning

Authentication Mobility management Radio resource control NAS-AS integration

L1/L2 functions High CF with M-MIMO - for capacity

Extraction of actionable insights from big data Orchestration of required services and functionalities (e.g., traffic optimization, context-aware QoE provisioning, caching, ...)

API

XaaS

API

Operatorservices

Internet

CPE C-plane entity

C-plane path

UPE

UPE

NFV enabled NW cloud

Macro cell

Small cell

RRU

CPE

NI

U-plane entityNI Network intelligence

U-plane pathRadio access linkBackhaul (fiber, copper, cable)

Wireless fronthaul

AS Access stratumCF Carrier frequencyD2D Device-to-deviceM-MIMO Massive MIMOMTC Machine-type communicationsNAS Non-access stratumNOMA Non-orthogonal multiple accessNW NetworkOTA Over the airOTT Over-the-top playerRRU Remote radio unit

Data-drivenNW intelligence

C/U-plane split

Resource pooling

L1/L2 functions Low CF with NOMA - fall back for coverage High CF with M-MIMO - wireless backhaul

L1/L2 functions Super high CF and/or unlicensed spectrum - for local capacity Switched on on-demand

Dual connectivity Independent C/U-plane mobility

Nesting and relaying to support low-powered devices, nomadic cells and group mobility

Network-controlled D2D

Connectionless, contention-based access with new waveforms for MTC asynchronous access



Connectionless and contention-based accesswith new waveforms for asynchronousaccess of massive numbers of MTC devices

Data-driven network intelligence to opti-mize network resource usage and planning

LOGICAL NETWORK LAYERS: RADIO NETWORK AND NETWORK CLOUD

The network architecture consists of only twological layers: a radio network and a networkcloud. Different types of base stations and RRUsperforming a minimum set of L1/L2 functionsconstitute the radio network. The network cloudconsists of a U-plane entity (UPE) and a C-plane entity (CPE) that perform higher-layerfunctionalities related to the U- and C-plane,respectively (Fig. 2).

As shown in Fig. 3, the physical realization ofthe network cloud could be tailored to meet vari-ous performance targets. For example, instancesof UPEs and CPEs could be located close to basestations and RRUs to meet the needs of latency-critical services. To support latency-critical ser-vices, for example, it may be better to connectRRU3 to a small nearby data center (data center3) rather than a large data center farther away(data center 2). On the other hand, RRU1 may beconnected to a large data center located fartheraway (data center 2) rather than a nearby smalldata center (data center 1) if support for latency-critical services is not required. Such flexibilityallows the operator to deploy both large and smalldata centers to support specific service needs.

Such architecture simplifies the network andfacilitates quick, flexible deployment and man-agement. Base stations would become simpler

and consume less energy due to the reducedfunctionalities, thereby making dense deploy-ments affordable to deploy and operate [15, 16].Additionally, the network cloud allows forresource pooling, reducing overprovisioning andunderutilization of network resources.

DYNAMIC DEPLOYMENT AND SCALING OFNETWORK FUNCTIONS WITH SDN AND NFV

By employing SDN and NFV, CPE and UPEfunctions in the network cloud can be deployedquickly, orchestrated and scaled on demand. Forinstance, when a local data center is unable tocope with a flash crowd (e.g., due to a local disas-ter), additional capacity can be borrowed quicklyfrom other data centers. In addition, resourceswithin a data center can be quickly shifted tosupport popular applications simply by addingadditional instances of the required software.

Besides this application-level flexibility, theuse of a cloud infrastructure also provides flexi-bility with respect to the available raw processingcapacity. Spare cloud resources can be lent outwhen demand is low, whereas additionalresources can be rented through infrastructureas a service (IaaS) business models during peakhours. Furthermore, a broad range of as a ser-vice business models based on providing specif-ic network functionalities as a service (i.e.,XaaS) could also be envisioned. The completeor specific parts of the network could be provid-ed to customers (e.g., network operators, OTTplayers, enterprises) that have specific require-ments, for example in a mobile network as aservice or radio network as a service model.UPE/CPE/NI as a service models, where spe-

Figure 3. Realization of a 5G network cloud. The network cloud is a logical entity with physical realization that can be tailored tomeet specific needs.

Data center 4

Data center 1

Data center 2

Network intelligence(data collection,

analyses and controlover network entities)

RRU3

RRU2

RRU1

Macrocell

Small cell

NFV-enablednetwork cloud

Data center 3

CPE C-plane entity

C-plane path

UPE U-plane entity

RRU Remote radio unit

U-plane pathWireless fronthaul



cific core network functionalities (Fig. 2) of themobile network are provided a la carte as a ser-vice, could also be envisioned. Last but not least,parts of the platform could be rented out tothird parties like OTT players to enable the pro-vision of services and applications that requireextremely low latency to end users. Besides theXaaS business models that could be facilitated,the flexibility of a cloud, coupled with SDN andNFV technologies, also makes the network easi-er, faster, and cheaper to deploy and manage.

LEAN PROTOCOL STACKWith virtualization, interfaces between networkfunctionalities become interfaces between soft-ware. Two separate protocols for the C-planemay no longer be relevant if both NAS and ASprotocols can be virtualized. Under a unifiedcloud paradigm, the NAS and AS protocols canbe integrated into a single protocol, removingredundant functionality. In current LTE, forexample, the NAS ServiceRequest and RRCConnectionRequest messages are concatenated,but these could be merged into a single messagein a future cloud-based and virtualized network.Similarly, some procedures related to mobilitymanagement, session management, and securitycan potentially be removed. As an example, theconnection establishment procedure can be sig-nificantly simplified by requiring a handshakeonly between the peer entities of a single proto-col. This in turn will realize faster connectionestablishment. Bearer-based QoS managementcould also be replaced by simple IP marking,with proper mechanisms in place to prevent allpackets being marked with the highest QoS class.

Similarly for the U-plane, merging of func-tionalities in the RAN L2 and gateway function-alities in the current core network (CN) can beconsidered. Virtualization of the U-plane is gen-erally considered to be more difficult than thatof the C-plane due to the sheer volume of datato be processed. Virtualization of the RAN L2protocols can demand significant processingpower, as L2 protocols support various featuresthat are dynamic in nature, like dynamic trans-port block size (according to resource allocationand instantaneous radio condition), segmenta-tion and concatenation of packets, and hybridautomatic repeat request (ARQ). The radioscheduler functionality and advanced featureslike mMIMO require accurate channel stateinformation (CSI) to be effective. Hence, if suchfeatures are to be virtualized, CSI also needs tobe delivered to the virtualized entity, potentiallyimposing significant transport overhead. Howev-er, with sufficient advancements in technologyand careful selection of functionalities, some ofthe services provided by L2 can be feasible forvirtualization around 2020. In principle, thisallows the functionalities provided by differentRAN and CN protocols to be merged and a sin-gle U-plane entity to provide radio transport ser-vices and gateway functionalities. Nevertheless,careful study is needed to determine for whichlayers such integration can occur.

One feature that can be potentially removedfrom the U-plane stack is ciphering, since this isincreasingly implemented by transport layersecurity (TLS) over IP. Generally, E2E solutions

are more efficient than encrypting segmentsalong the path. However, E2E encryption impliesno traffic visibility along the path and makestraffic control in networks difficult. In manyoperator networks today, intelligent mechanismssuch as deep packet inspection (DPI) andcaching are used to optimize resource usage andimprove QoE. End-to-end encryption wouldmake these intelligent mechanisms dysfunction-al. As security of signals transmitted over the airis essential, due to the broadcast nature of radiosignals, where to terminate ciphering in the net-work is an important issue.

INDEPENDENT PROVISIONING OFCOVERAGE AND CAPACITY WITHC/U-PLANE SPLIT ARCHITECTURE

Coverage and capacity are provided indepen-dently in the RN with a C/U-plane split architec-ture. Macro and metro base stations providecoverage using licensed spectrum in lower fre-quency bands and existing cell sites, integrating,for example, NOMA and SIC to boost capacity[8].

Small cell base stations (e.g., Phantom cells[9]) and RRUs provide localized capacity using acombination of licensed and unlicensed spec-trum in low and high frequency bands. Thesecells are deployed indoors and at outdoorhotspots. Advanced schemes (e.g., mMIMO) arealso implemented in some RRUs and small cellsto boost capacity. Because of the highly variableuser and traffic distribution in small cells, theycan be put to sleep or switched off completelywhen they are not needed to save energy.Dynamically switching small cells on and off canprovide significant energy savings withoutdegrading network performance [16].

Separating coverage from capacity enablesindependent mobility of the C-plane and U-plane in areas with overlapping coverage ofmacro and small cell base stations. In effect, theC- and U-planes for a terminal can take differ-ent paths. This requires the terminal to supportconnectivity to multiple base stations at the sametime.

RELAYING AND NESTING TO SUPPORTMULTIPLE DEVICES, GROUP MOBILITY, AND

NOMADIC HOTSPOTS

Relays are used as a means to support groupmobility (e.g., terminals in a moving vehicle) andnomadic hotspots. In such scenarios, all trans-missions within the group are aggregated at oneor more entities (e.g., a small cell) and relayedto the network through a wireless backhaul thatconnects to the network cloud (Fig. 2). Deviceswith limited resources, such as low-poweredwearable devices, connect non-transparently tothe network through one or more devices thathave more resources (nesting, Fig. 2). By con-necting non-transparently, network paging pro-cedures can be used to initiate connections tosuch devices, thus reducing signaling traffic andpower consumption. Together, relaying and nest-ing provide support for a huge number of deviceswith diverse capabilities in a scalable and effi-cient manner.

Because of the highly

variable user and

traffic distribution in

small cells, they can

be put to sleep or

switched off com-

pletely when they

are not needed to

save energy. Dynami-

cally switching small

cells on and off can

provide significant

energy savings with-

out degrading net-

work performance.



DATA-DRIVEN NETWORK INTELLIGENCEThe architecture allows the network cloud to col-lect various types of user-centric, network-centric,and context-centric data. The network cloud usesintelligent algorithms to provide real-time insightsfor efficient resource management, mobility man-agement, local offload decisions (e.g., network-con-trolled D2D communications), QoE management,traffic routing, and context-aware service provision-ing (e.g., geocasting). Furthermore, the aggregateddata can provide useful input for network plan-ning. By providing application programming inter-faces (APIs) to the network cloud, the collecteddata can be used in various forms for useful public(e.g., urban planning) and commercial purposes.For example, the APIs can be used to facilitatenew businesses based on selling knowledge aboutnetwork conditions as a service to OTT players,which can allow them to provide consistent servicequality to end users.

ISSUESSeveral issues need to be addressed in order torealize the proposed network architecture inparticular, and 5G networks in general. Some ofthese issues are summarized in Fig. 4 and brieflydiscussed below.

One issue that must be addressed is how lega-cy networks will interface and interoperate withthe new network architecture. One could imaginea migration step where the legacy CN and RANare migrated to separate cloud platforms duringthe development phase of 5G (Fig. 4). In orderto avoid building parallel networks, it will beessential to specify interfaces and protocolsbetween entities in the legacy clouds and the newnetwork cloud to ensure interoperability.

Another issue is to determine the optimalphysical realization of the network cloud to meetperformance and cost targets. Whereas central-ization of resources could result in savings frompooling, it could also lead to performance bottle-necks, higher latency, and single points of fail-ure. Additional robustness measures will also beneeded to avoid devastating impact on serviceavailability if the central entity fails. Moreover,centralization could lead to the need for largerprocessing and transport capacity at the centralentity to process and transport the aggregatedtraffic, which could diminish the cost savingsachieved by pooling. On the other hand, dis-tributing resources could lead to performanceimprovements and reduced latency, but may becostly due to reduced pooling gains and anincreased number of data center locations atcorresponding higher operational expenses.Finding the right balance is an important issue.

Ultra-dense small cell deployments will beespecially useful for indoor and hotspot environ-ments. As shown in Fig. 5, different deploymentoptions have different implications for the net-work. In addition to spectrum, backhaul is alsoan important issue, especially for user deploy-ment. Local breakout may be required for moreefficient routing through the user-provisionedbackhaul. However, this has implications on thefunctionalities needed at the small cell base sta-tion. For instance, U-plane processing function-alities are needed to support local breakout.Additionally, support for local breakout makestraffic invisible to the network, which affectsintelligent QoE provisioning.

Besides the issues highlighted above, seam-less mobility provisioning among different typesof deployed local and wide-area technologies

Figure 4. Overview of issues that must be addressed to realize the 5G architecture vision.

RAN cloud

CN cloud

Current NW Future NW

Migration and required interfaces

Minimum functionalities needed

Traffic management OTA security provisioning Seamless mobility

Device/ID management Device discovery

Deployment and spectrum Discovery and measurement Network synchronization

Support V2X communications

Support cognitive mobile objects

Support potentially different air interfaces

Intermediate step

LegacyCN

LegacyRAN

MME

BBU

P/S-GW

Macro cell

Small cell

BBU Baseband unitCN Core networkGW GatewayMME Mobility mgt. entityOTA Over the airOTT Over-the-top playerP/S Packet/servingRAN Radio access networkRRU Remote radio unit

Internet

OTT

API

XaaS

API

Operatorservices

NFV enabled NW cloudCPE

CPE

RRU

UPE

NI

C-plane entity

C-plane path

UPE U-plane entityNi Network intelligence

U-plane pathRadio access linkBackhaul (fiber, copper, cable)Wireless backhaul

By connecting non-

transparently, net-

work paging

procedures can be

used to initiate con-

nections to such

devices, thus reduc-

ing signaling traffic

and power con-

sumption. Together,

relaying and nesting

provides support for

a huge number of

devices with diverse

capabilities in a scal-

able and efficient

manner.



with potentially different functionalities also hasto be addressed to improve the overall QoE forend users. Mechanisms to support simultaneoussessions and seamless session mobility across dif-ferent access networks will also be required tosupport consistent QoE for end users. Further-more, different types of edge networks will alsoneed to be integrated within the 5G networkarchitecture. For instance, the communicationneeds of cognitive mobile objects (robots2X,drones2X, etc.) will all need to be efficientlysupported and integrated in 5G networks. Final-ly, new paradigms of identity management andcharging will need to be developed for 5G, inparticular, to cope with the huge number ofdevices expected to be connected to the net-work, the diverse use cases, and different edgenetwork topologies.

INITIAL PROOF OF CONCEPTA real-time simulator is used to evaluate the sys-tem-level gains when some of the candidate 5Gtechnologies described in the previous sectionare introduced for downlink transmission. Specif-ically, the gains from the hybrid usage of macro-cells at lower frequency bands and small cells athigher frequency bands, together with mMIMO,are demonstrated.

Figure 6 shows the deployment environmentstudied, which consists of buildings, moving vehi-cles, users, macro base stations, and a densedeployment of small cell base stations. A seven-cell model is assumed with an inter-site distanceof 500 m. Each macrocell has three sectors, andeach sector has 30 outdoor users (i.e., penetrationloss = 0 dB). A 3 km/h user speed is assumed.Ray tracing is applied using the vertical planelaunch (VPL) method to emulate a real propaga-tion environment of a 750 m 750 m dense urbanarea in Shinjuku, Tokyo. The baseline system

consists of LTE-based macrocells using 20 MHzbandwidth at 2 GHz. Each macrocell uses twotransmit (Tx) antennas. An antenna gain of 14dBi and a total transmit power of 49 dBm areassumed for each macrocell base station. Forevaluating the gains of network densification andwideband transmission at higher frequency bands,12 small cells are deployed per sector. Each smallcell uses 1 GHz bandwidth at 20 GHz. The num-ber of Tx antennas per small cell is 64. An anten-na gain of 5 dBi and a total transmit power of 30dBm are assumed for each small cell base station.The number of receive antennas at the user ter-minal is 4 at both 2 GHz and 20 GHz.

For 2 4 MIMO transmission in macrocells,single-user MIMO is applied based on implicit

Figure 5. Small cell deployment options and issues.

Unlicensedspectrum

Pros Reduced cost (equip., deployment, operation)

Cons Lack of QoE guarantees

Issues Access control Mechanisms to ensure fair-play (definition and implementation of incentive-compatible spectrum etiquette) Coexistence with Wi-Fi, Bluetooth, etc. Impact of diverse backhaul types on advanced RA techniques (e.g., CoMP) Provisioning of over-the-air security

Pros Cell sites fully controlled by the operator Additional spectrum for operators to exploit

Cons Cost (equipment, deployment, operation) Lack of QoE guarantees

Issues Mechanisms to ensure fair-play (definition and implementation of incentive- compatible spectrum etiquette) Coexistence with Wi-Fi, Bluetooth, etc. Backhaul provisioning

Licensedspectrum

Pros Reduced cost (equip., deployment, operation)

Cons Additional operation costs to provide after- service customer support

Issues Regulatory issues Access control (public or private) Ensuring QoE, e.g., new mechanisms to control interference (e.g., low Tx power) Impact of diverse backhaul types on advanced RA techniques (e.g., CoMP) Provisioning of over-the-air security

User-deployed

Pros Cell sites fully controlled by the operator Easier to provide QoE Advanced resource allocation (RA) techniques become easier to realize

Cons Cost (equipment, deployment, operation) Limited spectrum Spectrum license fees

Issues Backhaul provisioning

Operator-deployed

Figure 6. The deployment environment of the 5G real-time simulator.

Macrocell w/3 sectors

Movingvehicles

BeamsUsers

Smallcells

Buildings

Color code of users and the ground indicates achievable user data rate:Blue: 010 Mb/s; Green: 10100 Mb/s; Yellow: 100500 Mb/s;Orange: 500 Mb/s1 Gb/s; Red: 110 Gb/s; Pink: over 10 Gb/s



CSI feedback using the LTE Release 8 code-book. For 64 4 mMIMO transmission in smallcells, the CSI of users is assumed to be perfectlyknown at the small cell base station side, andHermitian precoding is applied for multi-layertransmission. In order to improve both cell cov-erage (by beamforming gain) and spectrum effi-ciency (by spatial multiplexing gain) of smallcells, single-user MIMO and multi-user MIMOdynamic switching (up to 4 users) and rankadaptation (up to 4 layers/user) are introduced.Proportional fairness scheduling is applied toallocate frequency/time resources to users atmacrocells and small cells disjointly. Note thatno intercell interference coordination (ICIC) isapplied among either macrocells or small cells.

The performance of the candidate technolo-gies are shown in Fig. 7. Figure 7a illustrates thespectrum usage for macrocells and small cells. Itcan be seen that the power spectrum density(PSD) becomes lower as the spectrum band-

width is extended to 1 GHz for small cells. InFigs. 7b and 7c, the x-axis (time [subframe])refers to the number of subframes being pro-cessed and also the time in milliseconds (onesubframe = 1 ms). The system throughput persubframe of a 500 m 500 m area is shown inFig. 7b, which demonstrates that compared to amacro-only 3GPP Release 8 LTE deployment,around 1300 system throughput gains areachieved by a combination of dense deploymentof small cells, using large bandwidths at higherfrequency bands and employing mMIMO tech-niques at small cells. By simulating each of thecandidate 5G technologies above, we see the1300 system throughput gains as the combina-tion of almost 50 from bandwidth extensionfrom 20 MHz to 1 GHz, 4 from antenna densi-fication by adding 12 small cells per sector, andaround 6.5 from mMIMO by introducing 64 4 mMIMO with single-user MIMO and multi-user MIMO dynamic switching.

Figure 7. Spectrum usage and performance evaluation results of the 5G real-time simulator: a) powerspectrum density (dBm/Hz) vs. frequency (GHz); b) throughput (Mb/s) vs. time (subframe); c) classi-fied UE ratio vs. time (subframe).

(a) (b)

(c)

Color code for Figure 7c indicates the fraction of users able to achieve a particular range of data rate: Blue: 010 Mb/s; Green: 10100 Mb/s; Yellow: 100500 Mb/s; Orange: 500 Mb/s1 Gb/s;

Red: 110 Gb/s; Pink: over 10 Gb/s

New paradigms of

identity management

and charging will

need to be

developed for 5G,

in particular, to cope

with the huge

number of devices

expected to be

connected to the

network, the diverse

use cases and

different edge

network topologies.



Finally, Fig. 7c shows the classified UE ratio,which gives the fraction of users who are able toachieve a particular range of data rate. It can beseen from this that more than 90 percent ofusers are able to achieve data rates in excess of 1Gb/s (i.e., the red color zone expanded to below0.1) with such a network. These initial resultsdemonstrate the potential of network densifica-tion using small cells, bandwidth extension inhigher frequency bands, and mMIMO at smallcells to address the capacity and data rate chal-lenges of 5G networks.

CONCLUSIONSThe important challenges that must be addressedby 5G networks have been highlighted: highercapacity, higher data rate, lower E2E latency,massive device connectivity, reduced capital andoperation cost, and consistent QoE provisioning.A 5G architecture vision to address some of thosechallenges is presented and a two-layer architec-ture proposed, consisting of a radio network anda network cloud. The proposed architecture inte-grates various enablers such as small cells, mas-sive MIMO, C/U-plane split, NFV, and SDN. Themain concepts can be summarized as follows: Ultra-dense small cell deployments on

licensed and unlicensed spectrum, underC/U-plane split architecture, to addresscapacity and data rate challenges

NFV and SDN to provide flexible networkdeployment and operation, with integratedAS and NAS features

Intelligent use of network data to facilitateoptimal use of network resources for QoEprovisioning and planningInitial proof of concept investigations suggest

more than 1000 times throughput gains com-pared to a macro-only 3GPP Release 8 LTEdeployment are achievable by a combination ofdense deployment of small cells, using largebandwidths at higher frequency bands andemploying massive MIMO techniques at smallcells. Nevertheless, some of the componentshighlighted in the system concept have mutualconflicts when details are considered. Hence,how to balance the pros and cons of each aspectneeds to be carefully studied. Further investiga-tions are necessary, particularly in the followingareas: suitable techniques for use in small cellsin different frequency regimes; how to incorpo-rate small cells with NFV and SDN in a cost-effective manner; and intelligent algorithms thatbetter utilize the available network resources toprovide a consistent end-user QoE.

REFERENCES[1] Qualcomm, The 1000x Mobile Data Challenge, White

Paper, Nov 2013. [2] NSN, Signaling is Growing 50% Faster than Data Traf-

fic, White Paper, 2012. [3] METIS, Scenarios, Requirements and KPIs for 5G

Mobile and Wireless System (Deliverable D1.1), May2013.

[4] Advanced 5G Network Infrastructure for the FutureInternet Public Private Partnership in Horizon 2020,2013.

[5] Y. Kishiyama et al., Future Steps of LTE-A: Evolutiontoward Integration of Local Area and Wide Area Sys-tems, IEEE Wireless Commun., vol. 20, no. 1, 2013,pp. 1218.

[6] E. G. Larsson et al., Massive MIMO for Next Genera-tion Wireless Systems, May 2013.

[7] S. Gollakota, S. Perli, and D. Katabi, Interference Align-ment and Cancellation, ACM SIGCOMM Comp. Com-mun. Rev., 2009.

[8] A. Benjebbour et al., System-Level Performance ofDownlink NOMA for Future LTE Enhancements, IEEEGLOBECOM, 2013.

[9] H. Ishii, Y. Kishiyama, and H. Takahashi, A Novel Archi-tecture for LTE-B: C-plane/U-Plane Split and PhantomCell Concept, IEEE GLOBECOM Wksps., 2012.

[10] G. Fodor et al., Design Aspects of Network AssistedDevice-to-Device Communications, IEEE Commun.Mag., vol. 50, no. 3, 2012, pp. 17077.

[11] G. P. Fettweis, A 5G Wireless CommunicationsVision, Microwave J., Dec 2012.

[12] B. Farhang-Boroujeny, OFDM Versus Filter Bank Multi-carrier, IEEE Signal Proc. Mag., vol. 28, no. 3, May2011, pp. 92112.

[13] EARTH Project Work Package 2, Deliverable D2.1: Eco-nomic and Ecological Impact of ICT, https://www.ict-earth.eu/publications/deliverables/deliverables.html,2011.

[14] Network Functions Virtualisation - Introductory WhitePaper, SDN and OpenFlow World Congress, Darm-stadt, Germany, Oct 2012.

[15] EARTH Project Work Package 4, Deliverable D4.3:Final Report on Green Radio Technologies,https://www.ict-earth.eu/publications/deliverables/deliv-erables.html, 2012.

[16] E. Ternon et al., Database-Aided Energy Savings inNext Generation Dual Connectivity Heterogeneous Net-works, IEEE WCNC, Istanbul, Turkey, Apr 2014.

BIOGRAPHIESPATRICK AGYAPONG is a researcher with the wireless researchgroup at DOCOMO Communications Laboratories Europe.His research focuses on designing incentive-compatiblealgorithms, protocols, and architectures to support nextgeneration mobile communication needs, spanning theareas of resource management, content distribution, net-work architecture, and business strategy. He holds a Ph.D.in engineering and public policy from Carnegie Mellon Uni-versity, Pittsburgh, Pennsylvania. He also holds an M.Sc. inelectrical engineering and a B.Sc. in electrical engineeringand computer science, both from Jacobs University Bre-men, Germany.

MIKIO IWAMURA is a director of the wireless research group atDOCOMO Communications Laboratories Europe. He receivedhis Ph.D. and M.Sc. degrees from Kings College London in2006 and the Science University of Tokyo in 1998, respec-tively. Before his current role, he was deeply engaged in LTEstandardization, with over 300 contributions to 3GPP. Hehas over 100 patents internationally, and has published over20 technical journals and conference papers.

DIRK STAEHLE is a manager at DOCOMO CommunicationsLaboratories Europe. He is responsible for the standardiza-tion team contributing to the 3GPP SA2 and ETSI NFV stan-dardization groups. He received his Ph.D. from theUniversity of Wrzburg in 2004 and continued as an assis-tant professor, leading the wireless network researchgroup before joining DOCOMO in 2011. His research activi-ties include NFV, machine-type communication, applicationand QoE-aware traffic and resource management, andradio network planning.

WOLFGANG KIESS studied at the Universities of Mannheimand Nice, and holds a diploma in business informationsystems from the University of Mannheim and a Ph.D. incomputer science from the University of Dsseldorf. He isthe leader of the virtualization research team at DOCO-MO Communications Laboratories Europe, focusing oncellular core network virtualization, cloud computing,and 5G.

ANASS BENJEBBOUR [SM] obtained his Ph.D. and M.Sc. intelecommunications in 2004 and 2001, respectively, andhis Diploma in electrical engineering in 1999, all fromKyoto University, Japan. He is currently an assistant man-ager of the 5G team within NTT DOCOMO, Inc. He servedas 3GPP RAN1 standardization delegate during LTE Release11, as secretary of the IEICE RCS conference, and Editorfor IEICE Communications Magazine. He is a Senior Mem-ber of IEICE.

Initial proof of con-

cept investigations

suggest more than

1000 times through-

put gains compared

to a macro-only

3GPP Release 8 LTE

deployment are

achievable by a com-

bination of dense

deployment of small

cells, using large

bandwidths at higher

frequency bands and

employing massive

MIMO techniques at

small cells.


/ColorImageDict > /JPEG2000ColorACSImageDict > /JPEG2000ColorImageDict > /AntiAliasGrayImages false /CropGrayImages false /GrayImageMinResolution 300 /GrayImageMinResolutionPolicy /OK /DownsampleGrayImages false /GrayImageDownsampleType /Average /GrayImageResolution 300 /GrayImageDepth -1 /GrayImageMinDownsampleDepth 2 /GrayImageDownsampleThreshold 1.50000 /EncodeGrayImages true /GrayImageFilter /DCTEncode /AutoFilterGrayImages true /GrayImageAutoFilterStrategy /JPEG /GrayACSImageDict > /GrayImageDict > /JPEG2000GrayACSImageDict > /JPEG2000GrayImageDict > /AntiAliasMonoImages false /CropMonoImages false /MonoImageMinResolution 1200 /MonoImageMinResolutionPolicy /OK /DownsampleMonoImages false /MonoImageDownsampleType /Average /MonoImageResolution 1200 /MonoImageDepth -1 /MonoImageDownsampleThreshold 1.50000 /EncodeMonoImages true /MonoImageFilter /CCITTFaxEncode /MonoImageDict > /AllowPSXObjects false /CheckCompliance [ /None ] /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false /PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true /PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXOutputIntentProfile (None) /PDFXOutputConditionIdentifier (CGATS TR 001) /PDFXOutputCondition () /PDFXRegistryName (http://www.color.org) /PDFXTrapped /False

/CreateJDFFile false /Description > /Namespace [ (Adobe) (Common) (1.0) ] /OtherNamespaces [ > /FormElements false /GenerateStructure false /IncludeBookmarks false /IncludeHyperlinks false /IncludeInteractive false /IncludeLayers false /IncludeProfiles true /MarksOffset 6 /MarksWeight 0.250000 /MultimediaHandling /UseObjectSettings /Namespace [ (Adobe) (CreativeSuite) (2.0) ] /PDFXOutputIntentProfileSelector /UseName /PageMarksFile /RomanDefault /PreserveEditing true /UntaggedCMYKHandling /LeaveUntagged /UntaggedRGBHandling /LeaveUntagged /UseDocumentBleed false >> > ]>> setdistillerparams> setpagedevice

grupo 4 desizxcgn considerations for a 5g network architecture

Documents

generation5g mobile

overall5g network

g mobile network architecture

g networks

mobile device

design choices

design andevolution

fundamental design