Mobile Network Architecture Evolution toward 5G - IMDEA Networks ...
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
LTE-Advanced Pro
Mobile Network Architecture
Evolution toward 5G
Peter Rost, Albert Banchs, Ignacio Berberana, Markus Breitbach, Mark Doll, Heinz Droste, Christian Mannweiler, Miguel A. Puente,
Konstantinos Samdanis, and Bessem Sayadi
The authors discuss 3GPP Abstract further services such as small data services as
EPS mobile network well as machine type communication (MTC)
evolution as a whole, ana- As a chain is as strong as its weakest element, services. Meanwhile, cloud computing technolo-
so are the efficiency, flexibility, and robustness gies and cloud concepts have gained momentum
lyzing specific architecture of a mobile network, which relies on a range of not only from the information technology (IT)
properties that are critical different functional elements and mechanisms. perspective, but also within the telecom world.
in future 3GPP EPS releas- Indeed, the mobile network architecture needs Integrating cloud concepts into 3GPP EPS allows
es. In particular, they dis- particular attention when discussing the evolu- support for novel and emerging services. On the
cuss the evolution toward tion of 3GPP EPS because it is the architecture other hand, it requires novel architectural con-
that integrates the many different future tech- cepts, which natively support cloud technologies.
a “network of functions,” nologies into one mobile network. This article However, the static assignment of functionality
network slicing, and discusses 3GPP EPS mobile network evolution as to network elements and the strong functional
software-defined mobile a whole, analyzing specific architecture proper- dependencies within each network element make
network control, manage- ties that are critical in future 3GPP EPS releases. it difficult to support the required flexibility of
ment, and orchestration. In particular, this article discusses the evolution future 3GPP EPS deployments.
toward a “network of functions,” network slicing, The following sections detail concepts that
and software-defined mobile network control, could contribute to the evolution of 3GPP EPS
management, and orchestration. Furthermore, in order to provide the required flexibility for
the roadmap for the future evolution of 3GPP supporting network services with diverse require-
EPS and its technology components is detailed ments, to enable diverse mobile networks deploy-
and relevant standards defining organizations are ments, and to provide a higher degree of context
listed. awareness. Specifically, the next section intro-
duces relevant concepts such as flexible function
Introduction composition, network slicing, and software-de-
The Third Generation Partnership Project fined network control. After that we provide an
(3GPP) evolved packet system (EPS) of Long overview of the standardization roadmap, and
Term Evolution (LTE) refers to the logical the article concludes in the final section.
architecture composed of the radio access net-
work (RAN), called the evolved universal ter- Mobile Network Evolution
restrial radio access network (E-UTRAN) in the In order to support diverse services such as
case of LTE, and the evolved packet core (EPC) eHealth, the Internet of Things (IoT), and
as defined in [1, 2] and illustrated in Fig. 1. The vehicular-to-everything (V2X) in future mobile
objective of this logical architecture is to enable a networks, we see a need for enhancing the
flat IP-based network and provide a standardized EPS toward a flexible mobile network accom-
set of network elements and network interfac- modating novel architectural principles while
es. Standardized elements and interfaces enable maintaining backward compatibility. Such an
operators to integrate equipment and implemen- evolved EPS architecture must support legacy
tations from different vendors into a single sys- radio technologies as well as novel radio access
tem, while ensuring interoperability. The design interfaces such as millimeter-wave (mmWave) or
of a logical architecture satisfies requirements centimeter-wave transmission. It should accom-
originating from use cases that are expected to modate emerging processing paradigms such as
be of particular interest for 3GPP EPS. So far, mobile edge computing (MEC) and cloud-RAN
the aim of 3GPP EPS has been mainly the pro- (C-RAN), while enabling flexible deployment
vision of mobile broadband service, for which patterns based on small, micro, and macrocells
the system makes very efficient use of available and allowing programmability to support very dif-
spectrum. ferent requirements in terms of latency, robust-
So far, past releases (i.e., Rel-11, Re-12, and ness, and throughput.
Rel-13) studied and specified how to integrate Based on this, we see two main objectives
Peter Rost and Christian Mannweiler are with Nokia Networks; Albert Banchs is with IMDEA Networks; Ignacio Berberana is with Telefonia I+D;
Markus Breitbach and Heinz Droste are with Deutsche Telekom; Mark Doll was with Alcatel Lucent and is now with Nokia; Miguel A. Puente is with ATOS;
Konstantinos Samdanis is with NEC Europe Labs, UK; Bessem Sayadi was with Alcatel Lucent, France, and is now with Nokia.
84 0163-6804/16/$25.00 © 2016 IEEE IEEE Communications Magazine • May 2016that must be addressed by an evolved 3GPP EPS
architecture.
Operator’s IP services
Multi-service and context-aware adaptation
of the mobile network, which implies that the
mobile network needs to adopt its operation Control layer Rx SGi Data layer
based on the actual service requirements and
the related context. The context includes deploy- Gx
ment properties, transport network properties, HSS PCRF PDN-GW
and service properties, as well as available RAN
technologies. S6a S5
Mobile network multi-tenancy, which aims to
reduce capital and operational costs by allow- S11
ing infrastructure providers to make the best MME S-GW
use of available resources, including spectrum
S1- 1-U
and infrastructure. Hence, multiple tenants may S1-MME MM S S1-U
E EPC
share resources within the mobile network while
offering diverse services. E-UTRAN
In order to achieve these objectives, the fol-
lowing main functionalities should be supported
X2
and will be further detailed in the following sec-
tions. eNB eNB
LTE-Uu LTE-Uu LTE-Uu
Network of functions: Traditionally, mobile
network functions are readily grouped into net-
work entities, each responsible for a predefined
set of functions, and interfaces connecting these
entities. Using a flexible “network of functions”
allows adaptation to diverse services, and optimi- Figure 1. The (basic) 3GPP evolved packet system.
zation using different software rather than using
different parameterizations. Each block may be
replaceable and could be individually instantiat- limited. For instance, it is possible to collocate
ed for each logical network running on the same EPC elements, such as gateways, with a base
infrastructure. However, it must not imply a mul- station in 3GPP EPS. However, it is not possi-
titude of interfaces, as detailed later. ble to only place parts of the functionality of a
Network slicing allows the same mobile net- gateway or mobility management entity (MME)
work infrastructure to be used by multiple differ- with a base station. Similarly, it is possible to
ent operators, including vertical market players, fully centralize RAN functionality using the com-
each implementing its own logical network, for mon public radio interface (CPRI) and central
example, a logical network for mobile broadband baseband units. However, such deployments
with very high throughput, a logical network use non-virtualized baseband units at the cen-
connecting a massive amount of sensor nodes tral location; hence, it is rather relocating func-
(including indoors), or a logical network provid- tionality that does not exploit all characteristics
ing critical infrastructure connectivity for traf- of cloud computing. It is further not possible to
fic management or energy control. Hence, each only move parts of the RAN functionality except
network slice fulfills different requirements and in a proprietary way [3, 4].
serves very different purposes. The decomposition of the mobile network
Software-defined mobile network control is functionality would imply a stronger decou-
required to flexibly control both a flexible net- pling of logical and physical architecture than in
work of functions as well as a set of network slic- 3GPP EPS as illustrated in Fig. 2, that is, phys-
es. This control must be programmable in order ical network functions (PNFs) may be executed
to adapt the network behavior to the current on bare metal, while virtual network functions
requirements. This functionality goes beyond the (VNFs) may be executed on local or remote data
separation of the control and data planes, includ- centers (referred to as edge and central cloud
ing the control of RAN functionality as well as in Fig. 2). Bare metal refers in this case to the
the mobile network control plane. non-virtualized access to radio access resourc-
es, for example, through digital signal proces-
Network of Functions sors (DSPs), rather than on cloud computing
The objective of a mobile network architecture platforms. Hence, depending on the use case,
is to allow for integrating different technologies requirements, and the physical properties of the
and enabling different use cases. Due to the part- existing deployment, mobile network functional-
ly conflicting requirements, it is necessary to use ity is executed at different entities within the net-
the right functionality at the right place and time work. This imposes a number of challenges; for
within the network. In order to provide this flex- example, the system itself must not become more
ibility, it has recently been discussed whether the complex, and the introduction of new interfaces
network functions virtualization (NFV) paradigm should be avoided as much as possible. Hence,
should be adopted in the mobile access network the VNF assignment should exploit an effi-
domain, that is, enabling mobile network func- cient control and orchestration plane as further
tionality to be decomposed into smaller function described below. Furthermore, the coexistence
blocks that are flexibly instantiated. of different use cases and services would imply
So far, the degrees of freedom for assigning the need to use different VNF allocations with-
network functionality to network entities is very in the network. This is further elaborated later
IEEE Communications Magazine • May 2016 85Network slicing is cen-
Edge cloud
tered on the concept
Edge cloud Orchestration and
of deploying multiple management
VNF
dedicated logical mobile
VNF Controllers
networks with varying
levels of mutual isola- Bare metal
tion on top of the same
infrastructure. A network
PNF SDN-enabled transport
slice is a collection of network
Central cloud
mobile network func- Orchestration and
tions and a specific set management
VNF
of radio access tech- Bare metal Controllers
SDN-enabled transport
nologies necessary to network
operate an end-to-end
PNF
logical mobile network.
Figure 2. Relationship of functional assignment and physical architecture.
using the network slicing model. The challenge access technologies (RATs) (or specific RAT
of avoiding many additional interfaces may be configurations) necessary to operate an end-
addressed by a flexible container protocol on the to-end (self-contained) logical mobile network.
user [5] and control planes. The mobile network This set of network functions and configurations
must further integrate legacy technologies as well may be combined such that slice-specific data
to guarantee that it can operate with existing net- and control plane functionality is tailored to the
works. requirements of considerably different use cases,
The main benefit of the described architec- network customers, or business models. Con-
ture is the possibility to exploit centralization sequently, network slicing is a technology that
gains where possible, to optimize the network enables both multi-tenancy and service-tailored
operation to the actual network topology and composition of mobile networks.
its structural properties, and to use algorithms Network slicing leverages the economies of
optimized for particular services, that is, optimize scale to be expected when running multiple log-
through dedicated implementations instead of ical mobile networks on top of a common infra-
parameters. structure. In this sense, network slicing is an
Table 1 lists examples where the operation evolution of network sharing, which has been a
may be optimized through different VNFs. For key business model for mobile network operators
instance, it may be possible to use a flexible air to reduce deployment and operational costs. In
interface numerology and, depending on the 3GPP, the System Architecture 1 working group
network terminal, different coding strategies, (WG SA1) conducted a study on actively shar-
multiple-input multiple-output (MIMO) modes, ing RAN resources while maintaining sharing
and framing structures, which are optimized for policies and providing flexibility for on-demand
throughput, delay, or reliability. However, the resource sharing within shorter time periods [6].
upper layer packetization may still be the same Architecture and operations that enable differ-
for all use cases, which allows the same software ent mobile operators with a separate core net-
implementation to be reused. Another example work (multi-operator core network, MOCN) to
includes cooperative transmission, where gains share the RAN are specified by WG SA2 [7]. In
are highly dependent on the environment; for general, sharing of resources can be divided into
example, if the system is not operating at full three categories: static [8], dynamic (e.g., spec-
load, cooperative scheduling may perform as effi- trum sharing [9]), and mixed resource allocation
ciently as cooperative multipoint transmission, (spectrum sharing and virtualized resource block
whereas at full load the gains depend highly on sharing [10]). While passive and active sharing
the number of interferers and channel knowl- solutions, for example, for network elements
edge. or medium access control (MAC) schedulers,
are partially used and standardized today, these
network slIcIng sharing concepts are based on fixed contractual
Network slicing is centered on the concept of agreements with mobile virtual network oper-
deploying multiple dedicated logical mobile net- ators (MVNOs) on a coarse granularity basis
works with varying levels of mutual isolation on (monthly/yearly) [11].
top of the same infrastructure. A network slice NFV, and software-defined mobile network
is a collection of mobile network functions (or control and orchestration enable a new level of
groups of functions) and a specific set of radio sharing by decoupling infrastructure resources
86 IEEE Communications Magazine • May 2016from application software, and by a split of the
Network functions Relevant parameters
control and data planes. This significantly sim-
plifies the partitioning of network infrastructure
Highly depends on carrier frequency (e.g., sub-6 GHz or mmWave),
resources among different operators (or ten- Cell discovery MIMO technologies (e.g., beamforming).
ants). Further, slices can be isolated from each
other to allow for an adaptation of security mea- Mobility may not be required by some services (metering), or only
sures according to service-specific requirements Mobility very locally (enterprises), in groups (trains), or at very high speed
(flexible security) and for securing parallel oper- (cars).
ation of multiple services or tenants. While isola-
tion between network slices is highly important, Carrier aggregation may not be needed in each scenario as it also
Carrier aggregation impacts battery consumption; it could further include very distinct
it finds its limits where available resources need spectrum.
a common control (e.g., the radio scheduler): If
the required isolation level cannot be preserved,
a security weakness in one slice can be exploited Multi-connectivity could include different network layers (micro/
macro), different technologies (WiFi/LTE), and different spectrum
to attack another slice. Strong security measures Multi-connectivity (sub-6 GHz/mmWave). It may further be implemented at very dif-
to maintain the isolation between multiple ser- ferent layers (e.g., among others) depending on deployments.
vices and tenants operating on a shared infra-
structure platform must be mandatory for all The actual connectivity may be based on bearers (high throughput)
services and tenants. Connectivity model or connectionless (IoT). In the connectionless case, many non-ac-
Mobile core network elements rapidly evolve cess stratum (NAS) functions are not needed.
toward “cloud readiness” (i.e., deployment in
data center environments). Consequently, each Coding techniques may vary depending on the use case, for exam-
network slice can be composed from dedicated, Coding ple, block codes for short (sensor) transmissions or turbo codes for
high throughput.
customized instances of required network func-
tions (NFs) and network elements (NEs). Alter-
natively, slices can share function instances in Depending on the current load, deployment, and channels, tighter
Multi-cell cooperation cooperation (joint Tx/Rx) or looser cooperation (ICIC) is possible.
particular cases (e.g., for storage-intensive com-
ponents like subscriber databases). In the RAN
Depending on the use case requirements and available spectrum,
domain, extended sharing concepts facilitate the Spectrum access possibly different spectrum access strategies may be required (e.g.,
exploitation and management of radio resources licensed, unlicensed, license-assisted).
offered by the owner of the network infrastruc-
ture to tenants. In this multi-tenant ecosystem, Depending on the applicable access control and accounting/
Authentication, authorization,
classic tenants such as mobile network operators charging policies, AAA functionality is different and may be placed/
accounting (AAA)
(MNOs) and mobile virtual network operators instantiated in different locations.
(MVNOs) coexist with vertical businesses, for
Depending on the user context (children) and the requested ser-
example, utility companies, automotive and man- Parental control vice, the parental control function becomes part of the service chain
ufacturing companies, and over-the-top (OTT) for according service flows.
service providers such as YouTube and Netflix.
These tenants relate to network slicing in the Table 1. Examples for functional optimization.
sense that a tenant may instantiate and make use
of one or more slices. Figure 3 shows how the
different NFs may be instantiated on different that context, software-defined network (SDN)
network elements depending on the network functionality has recently gained momentum as
slice (service), that is, physical NFs would be a new approach to performing network opera-
deployed on non-virtualized hardware, different tions. With traditional SDN, control functions
levels of edge cloud instances would provide vir- are decoupled from the data plane through a
tualized resources (e.g., closer to the access point well defined interface and are implemented in
or exploiting points of presence) in addition to software. This simplifies networking, provides a
a central cloud. It further shows the virtualiza- higher degree of flexibility and enhanced scal-
tion layer, which is responsible for multiplexing ability, while reducing cost. Indeed, by simply
requests from different slices operating on virtu- modifying the software of the control functions,
alized resources toward physical resources. SDN allows the behavior of the network to be
Beyond multi-tenancy, network slicing addi- flexibly changed, considering specific services and
tionally serves as a means to deploy multiple ser- applications.
vice-tailored mobile network instances within a Following the paradigm of SDN, the con-
single MNO, each addressing a particular use case trol of the mobile network architecture adopts
with a specific set of requirements (e.g., mobile the software-defined mobile network control
broadband or IoT). In that context, the aforemen- (SDMC) concept focusing on wireless-specific
tioned “network of functions” concept enables the functions. Our SDMC approach resembles SDN
joint optimization of mobile access and core net- by splitting wireless functionality into those func-
work functions. Each network slice is composed of tions that are being controlled and remain rela-
functions according to service needs; for example, tively stable, and those functions that control the
low-latency services require the allocation of most overall network and are executed at the control-
network functions at the edge. ler. However, our SDMC concept is specifically
devised to control mobile network functionality,
Orchestration and Management and it is not limited to data plane functions, but
As mentioned before, an essential component includes control plane functions of the mobile
of the mobile network is the efficient orchestra- network, both of which can be placed arbitrarily
tion and management of mobile network func- in the edge cloud or the central cloud, as shown
tions through a low-complexity interface. In in Fig. 2.
IEEE Communications Magazine • May 2016 87Adopting a logically cen-
tralized control unifies
heterogeneous network SDM control and orchestration
Network slice 3
Network slice 2
technologies and pro- NF 1
Network slice 1
vides efficient network NF 2
control of heteroge-
NF 3
neously deployed net-
NF 4
works. In particular, the
network control must NF 5
consider evolving traffic Tenant 1
Compute Compute Compute
Virtualized
resources
demands, enhanced
Storage Storage Storage Tenant 2
mobility management, Networking Networking Networking
and dynamic radio char-
Virtualization layer
acteristics.
Compute Compute Compute
resources
Dedicated
Physical
special Storage Storage Storage
hardware Networking Networking Networking
Physical Edge Edge Central
NF cloud cloud cloud
Figure 3. Network slicing concept.
To enable the SDMC paradigm within 3GPP ager needs to interact with 3GPP policies, that is,
EPS, where wireless functionality is controlled the policy and charging rules function (PCRF),
centrally, we collocate the SDMC within the via a new network interface called ItfPolicy, to
3GPP network management system. This takes enable flexible policy provision for multiple ten-
advantage of the legacy performance monitor- ants and network innovation.
ing, forming a logical global RAN information The key advantages resulting from the pro-
base that can be used by the SDMC to control posed approach include the following.
various network functions. The control of wire- Flexibility: One of the problems that net-
less networks comprises, among others, chan- work operators are facing today is that while
nel selection, scheduling, modulation and coding wireless equipment is quite expensive, this is
scheme selection, and power control. Figure 4 very rigid and does not adapt to their needs. By
illustrates the SDMC architecture showing the using SDMC, operators would be able to fit the
main functional features and operations. With equipment to their needs through simply repro-
a software-defined approach, all these functions gramming the controller and thus reducing costs,
could be performed by a programmable soft- while being able to scale up and down virtual
ware defined mobile controller, which provides functions, also enhancing reliability.
very important benefits for the operation of the Unified Management: Adopting logically
mobile network. centralized control unifies heterogeneous net-
However, it is essential to enhance the cur- work technologies and provides efficient network
rent 3GPP Type 2 interfaces (ItfN) between the control of heterogeneously deployed networks.
network management system and the network In particular, the network control must consid-
equipment to allow the SDMC to provide net- er evolving traffic demands, enhanced mobility
work programmability and support for multi-ten- management, and dynamic radio characteristics.
ancy. Those enhancements should reflect SDN Simplified Operation of the Wireless Net-
capabilities such as network abstraction and con- work: With SDMC, network operators only need
trol providing sufficient network management to control a set of logically centralized entities
flexibility. Interfacing the SDMC with the net- that run the entire network, which, depending
work management system in such a manner can on actual latency requirements, possibly includes
also enable multi-tenancy support and network heterogeneous radio technologies.
programmability taking advantage of the 3GPP Enabling Network Innovation: By modify-
Type 5 interface. This allows receiving network ing the controller functions (i.e., SDMC Apps),
sharing requests from MVNOs [12] and offering many new services that were not included in
a means of network resource acquisition to OTT the initial architecture design can be enabled by
providers and verticals via the SDMC north- modifying the network behavior to introduce ser-
bound application programming interface (API). vice-specific enhancements within a few hours
In addition, the northbound interface offers the instead of weeks [13].
capability of flexible provision of the so-called Programmability: By adapting the functions
SDMC Apps. To accommodate the related ser- such as scheduling or channel selection to the
vice requirements of multi-tenancy and SDMC specific needs of the applications or the scenario,
Apps, the infrastructure provider network man- significant performance gains can be achieved.
88 IEEE Communications Magazine • May 2016By adapting the func-
Multi-tenancy support
tions such as scheduling
Updates QoS Innovation or channel selection to
the specific needs of
Algorithms Orchestration the applications or the
Sharing operator
SDMC apps Tenant A Tenant B network manager scenario, significant per-
Tenant C formance gains can be
Northbound-API (SDN) achieved. For instance,
Type 5 interface (3GPP)
Power efficiency Software defined mobile network controller (SDMC) the controller has a
scheduling
network slicing Itf-policy global view of the net-
RAN Infrastructure provider network manager
etc. info-base work, which allows for
Type 2 interface (Itf-N) (3GPP) optimizing the resource
with SDN related enhancements PCRF
allocation and schedul-
ing across multiple BSs.
X2
X2 X2
Shared-RAN
Figure 4. SDMC architecture and operations.
For instance, the controller has a global view • It is anticipated that the timeframe for pro-
of the network, which allows for optimizing the posals will be focused on 2018.
resource allocation and scheduling across multi- • In 2018–2020 the evaluation by independent
ple BSs. external evaluation groups and definition of
Inter-Slice Resource Control: Following the the new radio interfaces to be included in
network slice concept described above, infra- IMT-2020 will take place.
structure domain-hosted SDMC allows the infra- Similar to previous mobile network genera-
structure provider to assign unutilized resources tions, 3GPP is expected to also be the leading
to support third party services. Hence, the SDMC standardization body for 5G, and the correspond-
can allocate a network slice with a specified net- ing roadmap is shown in Fig. 5. 3GPP has start-
work capacity, a particular split of the control/ ed to work on 5G in both the SA and RAN
data plane, and a selection of VNFs. working groups. The current 3GPP Release 13
and the coming 3GPP Release 14 will provide
stAndArdIZAtIon roAdMAP enhancements to LTE-Advanced under the
The International Telecommunication Union name “LTE-Advanced Pro.” This will become
Radiocommunication Standardization Sector the baseline technology for the evolution from
(ITU-R) is developing a longer-term vision of LTE-Advanced to 5G. In parallel, 5G scenarios
mobile networks and their evolution toward 2020 and requirements will be studied, which likely
and beyond. It provides a framework and over- demand a revolutionary new architecture pro-
all objectives of the future developments of 5G viding greater flexibility, as stated in the previous
systems (referred to as IMT-2020) which involve section. This work is expected to be completed
several steps: by mid-2017.
• In early 2012, ITU-R embarked on a pro- SA1 has been working on a “Study on New
gram to develop “IMT for 2020 and Services and Markets Technology Enablers”
beyond,” setting the stage for fifth gener- (SMARTER) since April 2015. As a result, four
ation (5G) research activities, which are additional study items have been created that
emerging around the world. include three vertical industries and one hori-
• In 2015, ITU-R finalized its vision of the 5G zontal group. The verticals are enhanced mobile
mobile broadband connected society, which broadband (eMBB), critical communications
will be instrumental in setting the agenda (CriC), and massive IoT (mIoT); the horizontal
for the World Radiocommunication Con- study is on network operation (NEO). The latter
ference 2019, where deliberations on addi- deals with, among other issues, network slicing,
tional spectrum will take place in support of interworking, and migration, as well as fixed-mo-
the future growth of IMT. bile convergence (FMC). In March 2016, another
• In the 2016–2017 timeframe, ITU-R will study item for 5G vehicular-to-anything (V2X)
define in detail the performance require- communication was agreed. SA1 plans to finalize
ments, evaluation criteria, and methodolo- its studies in June 2016 and then start normative
gy for the assessment of a new IMT radio work in 3GPP Release 15.
interface. SA2 targets to finish its “Study on Architec-
IEEE Communications Magazine • May 2016 89The mobile network 2016.3 2017.6 2018.9 2019.12
Rel-15 Rel-16
architecture evolution as Rel-14 (initial 5G) (comprehensive 5G)
discussed in this article 5G SI(s): Scenario and requirements ...
impacts many different SI: HF (> 6GHz) channel model
network components.
Other features/enhancements
Hence, in addition to 5G SI(s): New RAT framework High-frequency-specific
3GPP other standards feature design
developing organiza- 5G phase 1 WI(s) 5G phase 2 WI(s)
Start with restricted spectrum
tions will participate in range Fundamental features of new RAT: Extension to full spectrum range,
restricted spectrum range, support other features and enhancements.
the definition of the both eMBB and IoT Meet all IMT-2020 requirements
future mobile network
ITU process Proposal Spec
architecture.
Figure 5. 3GPP LTE standardization roadmap toward 5G.
ture for Next Generation System” in September tions (SDOs) will participate in the definition
2016. An important topic in this study will be of the future mobile network architecture. Most
the interface between the LTE-Advanced RAN notably, the following SDOs will be involved in
and a future 5G core network (CN). SA2 has addition to 3GPP:
agreed to follow the Next Generation Mobile •The European Telecommunications Stan-
Network (NGMN) alliance, in particular Option dards Institute (ETSI) NFV industry specification
3 detailed in [12]. The new 5G CN will be able group (ISG) has created a framework for virtual-
to support a new 5G RAT as well as an evolved ization of network functions. This framework has
LTE-Advanced and other RATs such as IEEE been applied successfully to VNFs, mostly in the
802.11. This enables 5G network terminals to CN. In the RAN, where hardware still plays an
move between 5G and the evolved LTE-Ad- important role, implementation of NFV concepts
vanced without any interworking between the 5G is more difficult [14]; for example, the C-RAN
and 4G CNs, and thus provides a sound migra- concept with a fully centralized and virtualized
tion path from the LTE-based RAN to 5G. RAN was among the first use cases, already dis-
The RAN working groups are targeting the cussed in 2012 in ETSI NFV. However, as of
first true 5G features to appear in 3GPP Release today, there are no large-scale commercial imple-
15 (i.e., in the second half of 2018). This implies mentations. In order to gain more impact, the
that 3GPP will complete its initial 5G specifica- ETSI framework must be extended to be appli-
tions right before the Olympic Winter Games cable not only to virtualized hardware but also to
2018, which will take place in Korea. The focus non-virtualized, bare metal hardware [14].
in this 5G “Phase 1” will mostly be on enabling •The ETSI MEC ISG is looking at how to
new spectrum in high frequencies above 6 GHz. provide IT and cloud computing capabilities
More features for implementing architectural within the RAN in close proximity to mobile
enhancements will follow in 5G “Phase 2” with subscribers, allowing content, services, and
3GPP Release 16 (i.e., by the end of 2019) in applications to be accelerated, and increasing
time for their submission to the IMT-2020 as well responsiveness from the edge.
as the Olympic Summer Games 2020 in Japan. •The Open Networking Foundation (ONF)
Despite these planned architectural enhance- is the leading force in the development of open
ments, further efforts are needed by 3GPP and standards for the adoption of the SDN con-
other standardization bodies to accomplish the cept. However, in order to provide the benefits
migration from 3GPP EPS toward a new 5G described above, the SDN protocol functional-
architecture. A completely new type of inter- ities developed by ONF (e.g., OpenFlow and
face has to be designed and standardized when OF-Config) need to be extended to cope with 5G
the “network of functions” is going to replace requirements and toward 3GPP EPS.
today’s “network of entities,” as pointed out ear- •The Internet Engineering Task Force
lier. Furthermore, the use of network slicing for (IETF) is also considering the use of Internet
multi-tenancy and multi-service described above protocols (e.g., IPv6 and IP Multicast) in 5G net-
requires a flexible execution environment that works, although the work required does not have
is capable of supporting the diversity of network a clear scope yet. There are proposals for using
functions in parallel. The application of SDN IETF developed protocols such as locator/ID
concepts promising this flexibility to mobile radio separation protocol (LISP), host identity proto-
networks is, however, still in an experimental col (HIP), and information-centric networking
phase, although the C-RAN concept, RAN virtu- (ICN) to address shortcomings of the current 4G
alization, and their expected centralization gains CN for the support of additional 5G functional-
have been discussed for several years. ities (e.g., reducing network latency or support-
The mobile network architecture evolution ing new mobility models). IETF is also working
as discussed in this article impacts many differ- on the development of an architecture for ser-
ent network components. Hence, in addition to vice function chaining that includes the necessary
3GPP, other standards development organiza- protocols or protocol extensions for the nodes
90 IEEE Communications Magazine • May 2016that are involved in the implementation of ser- pean projects 5G-NORMA and METIS-II, and business unit projects on 5G
architecture. He serves as a member of the IEEE ComSoc GITC, VDE ITG It is in our opinion of
vice functions, as well as mechanisms for steering Expert Committee Information and System Theory, and as Executive Editor of
traffic through service functions. IEEE Transactions on Wireless Communications. high importance to
conclusIons And further chAllenges A LBERT B ANCHS [SM] (banchs@it.uc3m.es) received his M.Sc. and Ph.D. consider the future evo-
degrees from UPC-BarcelonaTech in 1997 and 2002. He was at ICSI Berkeley
This article discusses the evolutionary 3GPP EPS in 1997, at Telefonica I+D in 1998, and at NEC Europe from 1998 to 2003. lution of 3GPP EPS not
mobile network architecture, and the need to pro- Currently, he is an associate professor with the University Carlos III of Madrid, only as the introduction
vide a flexible architecture that integrates differ- and has a double affiliation as deputy director of the IMDEA Networks insti-
ent technologies and enables diverse use cases. tute. His research interests include performance evaluation and algorithm of a novel air interface
design in wireless networks.
We introduce and explain various concepts such but as the evolution of
as the transition from a predefined set of functions IGNACIO BERBERANA (ignacio.berberana@telefonica.com) received his M.S.
one mobile network
grouped into network entities to a flexible network degree in mining engineering from Madrid Polytechnic University in 1987. In
of functions, the network slicing concept, and soft- 1988 he joined Telefonica I+D, where he has worked mainly in wireless com- architecture toward a
munications, including several European projects (CODIT, MONET, Artist4G,
ware defined mobile network control, orchestra- iJOIN). Currently, he is responsible for the Innovation unit in the Radio Access “system of systems”
tion, and management. In addition, the relevance Networks direction of the Telefónica Global CTO office, dealing with long-
of different standards defining organizations has term evolution of mobile access, including 5G systems. where many different
been outlined and their roadmap has been detailed. use cases, technologies,
MARKUS BREITBACH (markus.breitbach@telekom.de) is working as a senior
It is in our opinion that it is highly important expert in the area of end-to-end network architecture. Before joining Deutsche and deployments are
to consider the future evolution of 3GPP EPS Telekom in 2006, he developed concepts for UMTS base stations and their
not only as the introduction of a novel air inter- HSPA schedulers for a major infrastructure supplier. In the last years, he has integrated, and the
face but as the evolution of one mobile network been working on network virtualization. Holding both a Ph.D. in electrical
engineering and an M.B.A., his ambition is to design innovative network operation of each
architecture toward a “system of systems” where concepts that fit well into the surrounding business picture.
many different use cases, technologies, and system is tailored to its
deployments are integrated, and the operation of MARK DOLL (mark.doll@alcatel-lucent.com) received his Dipl.-Phys. degree actual purpose.
each system is tailored to its actual purpose. in physics from Technische Universität Braunschweig in 2000 and his Dr.-
Ing. in computer science from Karlsruhe Institute of Technology (KIT) in
2007. At KIT, he worked on mobility, multicast, and QoS support for the
AcknowledgMent Internet. Upon joining Nokia Bell Labs, his work shifted to EPS CoMP and
This work has been performed in the framework EPS air-to-ground communication for aircrafts and now focuses on post-cel-
of the H2020-ICT-2014-2 project 5G NORMA. lular “user-centric” wireless access for 5G. He acts as 5G NORMA’s technical
manager.
The authors would like to acknowledge the con-
tributions of their colleagues. This information HEINZ DROSTE (Heinz.droste@telekom.de) works for Deutsche Telekom in
reflects the consortium’s view, but the consor- Darmstadt on mobile communication related projects. Antennas and radio
tium is not liable for any use that may be made wave propagation belong to his knowledge field as well as system-level
simulation and radio network planning. His current R&D activities at Telekom
of any of the information contained therein. Innovation Laboratories focus on the optimization of EPS and EPS-A deploy-
ments where he is acting as senior expert and project manager. He is actively
references contributing to the EU funded R&D project 5G NORMA.
[1] 3GPP, “TS 36.300; Overall Description; Stage 2,” tech. spec., 2013.
[2] 3GPP, “TS 23.401; General Packet Radio Service (GPRS) Enhancements for Christian Mannweiler (christian.mannweiler@nokia.com) received his
Evolved Universal Terrestrial Radio Access Network (E-UTRAN) Access,” M.Sc. (Dipl.-Ing.) and Ph.D. (Dr.-Ing.) degrees from Kaiserslautern Uni-
tech. spec., 2015. versity, Germany, in 2008 and 2014, respectively. Since 2015, he has
[3] P. Rost et al., “Cloud Technologies for Flexible 5G Radio Access Networks,” been a member of the Network Management Automation research group
IEEE Commun. Mag., May 2014. at Nokia. He has co-authored numerous articles and papers on future
[4] D. Wübben et al., “Benefits and Impact of Cloud Computing on 5G Signal mobile network technologies and architectures. He has worked in several
Processing,” IEEE Signal Processing Mag., Oct. 2014. nationally and EU-funded projects covering the development of cellular
[5] A. de la Oliva et al., “Xhaul: Toward an Integrated Fronthaul/Backhaul and industrial communication systems, among them H2020-5G-NOR-
Architecture in 5G Networks,” IEEE Wireless Commun., vol. 22, no. 5, MA, FP7-C-Cast, FP7-METIS, BMBF-SolarMesh, BMBF-PROWILAN, and
Oct. 2015. BMWi-CoCoS.
[6] 3GPP, “TR 22.852, Study on Radio Access Network (RAN) Sharing
Enhancements, Release 12,” tech. rep., June 2013. MIGUEL A. PUENTE (miguelangel.puente@atos.net) received his M.Sc. in tele-
[7] 3GPP, “TS 23.251, Network Sharing; Architecture and Functional Descrip- communications engineering from the Universidad Politécnica de Madrid
tion, Release 12,” Dec. 2013. (UPM) in 2012, including an information technology Master’s degree from
[8] S. Paul and S. Seshan, “Technical Document on Wireless Virtualization,” the University of Stuttgart (2010–2012). Since 2012 he is with Atos Research
GDD-06-17, GENI, Sept. 2006. & Innovation in Spain, where he is involved in European research projects
[9] Y. Zaki et al., “LTE Wireless Virtualization and Spectrum Management,” addressing 5G, EPS, Cloud Computing, Mobile Cloud/Edge Computing, QoE/
Proc. IEEE/IFIP Wireless and Mobile Networking Conf., Budapest, Hun- QoS optimization and recursive Internet architectures. From 2014 he is a
gary, Oct. 2010. Ph.D. candidate at UPM.
[10] T. Guo and R. Arnott, “Active RAN Sharing with partial Resource Reserva-
tion,” Proc. IEEE 78th VTC-Fall, Las Vegas, NV, Sept. 2013. KONSTANTINOS SAMDANIS (samdanis@neclab.eu) is a senior researcher and
[11] ITU-T Y.3011, “Framework of Network Virtualization for Future Networks,” backhaul standardization specialist with NEC Europe. He is involved in
Next Generation Networks-Future Networks, Jan. 2012. research for 5G networks and active in BBF on network virtualization, and
[12] NGMN Alliance, “NGMN 5G White Paper,” tech. rep., Feb. 2015. published numerous papers/patents. He has served as a Feature Topic
[13] C. J. Bernardos et al., “An Architecture for Software Defined Wireless Editor of IEEE Communications Magazine and IEEE MMTC E-Letters,
Networking,” IEEE Wireless Commun., vol. 21, no. 3, June 2014. Co-Chair of IEEE ICC 2014 and EuCNC 2015, and edited Green Commu-
[14] Small Cell Forum, “Network Aspects of Virtualized Small Cells, Release nications (Wiley). He received his Ph.D. and M.Sc. degrees from Kings
5.1, Document 161.05.1.01,” tech. rep., June 2015. College London.
BESSEM SAYADI (bessem.sayadi@alcatel-lucent.com) received his M.Sc. and
bIogrAPhIes Ph.D. degrees from SUPELEC in 2000 and 2003. He is a senior researcher at
PETER ROST [SM] (peter.rost@nokia.com) received his Ph.D. degree from Alcatel-Lucent Bell Labs. He has worked on several nationally and EU-funded
Technische Universität Dresden in 2009 and his M.Sc. degree from the Uni- projects. He has authored over 60 publications. He holds 20 patents and has
versity of Stuttgart in 2005. Since May 2015, he has been member of the more than 25 patent applications pending in the area of video coding and
Radio Systems research group at Nokia Germany, contributing to the Euro- wireless communications.
IEEE Communications Magazine • May 2016 91You can also read
