See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/342301961

A softwarized perspective of the 5G networks

Preprint · June 2020

CITATIONS READS
0 3,201

5 authors, including:

Cristiano Bonato Both Lucio Rene Prade

University of Vale do Rio dos Sinos (UNISINOS) Universidade do Vale do Rio dos Sinos
125 PUBLICATIONS 1,120 CITATIONS 27 PUBLICATIONS 20 CITATIONS

SEE PROFILE SEE PROFILE

Ciro J. A. Macedo Victor Hugo Lázaro Lopes

Instituto Federal de Educação, Ciência e Tecnologia de Goiás (IFG) Instituto Federal de Educação, Ciência e Tecnologia de Goiás (IFG)
10 PUBLICATIONS 7 CITATIONS 9 PUBLICATIONS 16 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

FUTEBOL - Federated Union of Telecommunications Research Facilities for an EU-Brazil Open Laboratory View project

Modeling and analysis of the Unisinos electrical system under aspects of power flow, short circuit, power quality, energy efficiency and distributed generation View project

All content following this page was uploaded by Cristiano Bonato Both on 19 June 2020.

The user has requested enhancement of the downloaded file.

 A softwarized perspective of the 5G networks
Kleber Vieira Cardoso, Cristiano Bonato Both, Lúcio Rene Prade,
Ciro J. A. Macedo, Victor Hugo L. Lopes

Abstract—The main goal of this article is to present the fundamental to clarify that this leap hides the various intermediate steps that
theoretical concepts for the tutorial presented in IEEE NetSoft 2020. comprise it. In the telecommunications industry, including for
arXiv:2006.10409v1 [cs.NI] 18 Jun 2020

The article explores the use of software in the 5G system composed of advertising reasons, a significant amount of technological devel-
the Radio Access Network (RAN) and the core components, following opments and innovations accumulate before defining a generation,
the standards defined by 3GPP, particularly the Release 15. The article
which was no different at 5G networks. As an example, the Re-
provides a brief overview of mobile cellular networks, including basic
concepts, operations, and evolution through the called ‘generations’ of
leases provided by the 3GPP (3rd Generation Partnership Project)
mobile networks. From a software perspective, RAN is presented in the that is one of the leading organizations for standardizing wireless
context of 4G and 5G networks, which includes the virtualization and mobile networks. The 4G/LTE networks (Long-Term Evolution)
disaggregation concepts. A significant part of the article is dedicated to were introduced in 2008 by Release 8 [5] and, since then, have
5G networks and beyond, focusing on core, i.e., considering the Service- received several updates (represented by new Releases) until the
Based Architecture (SBA), due to its relevance and totally softwarized 5G networks were officially introduced by Release 15 [6] in 2018.
approach. Finally, the article briefly describes the demonstrations pre- On the other hand, there are technical factors that also support the
sented in IEEE NetSoft 2020, providing the link for the repository that definition of each generation. In the 5G system, the radio spectrum
has all material employed in the tutorial.
usage is expanding to hundreds of MHz or some units of GHz,
but it also going to frequency bands of tens of GHz. Moreover,
5G networks consolidate an intense process of softwarization that
1 I NTRODUCTION stands out for the adoption of cloud systems and technologies
In the generations leading up to the fifth generation of wireless such as virtualization [7], software-defined networks [8], network
cellular mobile networks, the supreme effort was made in the slicing [9], and Service-Based Architecture (SBA) [10].
massification of access to customers and increasing the transmis- In addition to the technological aspects, the 5G networks in-
sion rates. However, more than just improving bandwidth and troduce changes in the companies’ business model. Traditionally,
reducing latency, 5G networks [1] will allow truly disruptive operators of wireless mobile communication systems have focused
solutions to emerging in all types of industries [2]. Although many on end-users as the main sources of revenue. In 5G networks, the
countries, such as Brazil, are still discussing and preparing for intention is to expand (or even migrate) the focus to have industries
the deployment of 5G networks, since 2019, several countries as primary customers [11], [12]. Three main requirements (or
(e.g., China, South Korea, and the United States) are already scenarios) have been defined to support this expansion of the
implementing, testing, and effectively using the latest generation telecommunications companies’ business model: enhanced mobile
of mobile wireless networks. More than 2.4 billion devices are broadband (eMBB), ultra-reliable low-latency communications
expected to use 5G networks by 2025 [3]. For now, rates close (URLLC), and massive machine-type communications (mMTC).
to (and even higher than) 1 Gbps are still the main novelties, These scenarios are illustrated in Fig. 1(a), which was introduced
but several services, mainly based on Internet of Things (IoT), by ITU (International Telecommunication Union) [13], in 2015, as
began to be planned and implemented. Autonomous cars, Industry part of the vision of what 5G networks would be. This illustration,
4.0, and virtual/augmented/mixed reality are some of the bets for which presents the leading capabilities of each scenario, is widely
relevant applications and tend to serve as interesting use cases for used to try to summarize some relevant properties of 5G networks.
new network features. However, as in previous generations, it is In order to meet the requirements of the different scenarios,
difficult to predict accurately which applications or services will the concepts of network software began to be adopted, both by
be adopted. For the time being, the only consensus is that 5G academia and industry, to make the development of 5G networks
networks, when deployed on a significant scale, will represent a (or systems) viable in recent years and also their evolution [9].
new step for wireless communications and will have a deep impact The main reason for the intensive introduction of software in
in the society [4]. 5G systems is to extend the flexibility to the mobile network
The fifth-generation represents a remarkable technological architecture, supporting different requirements in challenging sce-
leap over the fourth generation, introducing significant hardware narios emerging in the digitalized society. Moreover, with the
and, above all, considerable software innovations. It is essential network softwarization, it is possible to enjoy of traditional cloud
computing, as well as its smaller-scale variants, i.e., fog and
Kleber, Ciro, and Victor are with Universidade Federal de Goiás (UFG), edge. In this context, radio access and core network functions
Brazil. Email: {kleber, ciro.macedo, victor.lopes}@inf.ufg.br
Cristiano and Lúcio are with the University of Vale do Rio dos Sinos are now implemented using the advantages of cloud computing,
(UNISINOS), Brazil. Email: {cbboth, luciorp}@unisinos.br such as abundant storage, large-scale processing, elasticity, etc. In
 2

eMBB later, data-oriented communication devices. Since the first com-

Gigabytes in a second, data 3D video, UHD/4K/8K screens mercial mobile cellular network, in 1979, until nowadays, several
transfers
Gigabytesatem
extremely high
um segundo,
data rates
transferências de dados a taxa Vídeo
Work 3D, telas
and play in theUHD/4K/8K
Cloud
new technologies have been introduced in the communication sys-
de dados extremamente altas
Smart homes/buildings, video
tems, but also several concepts remain the same or are very similar.
Augmented reality
cameras
Domicílios/prédios inteligentes, In this section, we present relevant concepts of mobile networks in
Industrial automation,
câmeras de vídeo
Industry 4.0
general and the ones related to cellular networks. Additionally, we
Voice services describe the history of the mobile cellular networks very briefly
Mission-critical applications/
Serviços de voz through the ‘generations’, starting from the first until the fourth,
Smart cities, sensor services, e-Health
networks, IoT with most of the attention to the fourth generation. The fifth-
Autonomous cars
generation and beyond have the next section dedicated to them.
mMTC URLLC
(a) 5G scenarios
2.1 Basic concepts

eMBB Peak data User perceived The fundamental idea of a cellular network is the use of
High
rate relevance
data rate multiple low-power transceivers (transmitters+receivers) to offer
Traffic
Average communication to mobile devices. Actually, the power of these
Low Spectral transceivers is low compared with their predecessors, but it is very
capacity by
efficiency
area high compared to mobile devices. Traditionally, these transceivers
are mounted in a mast or tower, at an elevated, and have a control
Network unit. This configuration is named as a Base Station or BS, which
energy Mobility is responsible for covering an area using a band of frequencies.
efficiency Assuming the most traditional positioning in the center of the
coverage area, the area’s conceptual representation as a hexagon
Connection
mMTC URLLC was the appropriate choice because the distance
density Latency √ between the
(b) Capabilities of the scenarios
centers of all adjacent areas is the same (equal to 3R). Therefore,
a mobile cellular network can be represented as a collection of
Fig. 1. Scenarios (or requirements) defined for 5G networks (left) and the neighboring regions, each resembling a cell, as illustrated
capabilities offered in these scenarios (right). in Fig. 2(a). In practice, the environmental characteristics and
the propagation conditions tend to degenerate the mobile cellular
network into something more similar to a Voronoi diagram or
addition to enable new services, in the long run, the softwarization tessellation, as illustrated in Fig. 2(b).
is expected to reduce the costs of deploying and operating 5G The deployment of the BSs is generally defined by a previ-
systems. ous study that takes into consideration factors as topographical
The remainder of this article is structured as follows. Sec- characteristics, signal propagation conditions, limitations on siting
tion 2 introduces basic concepts related to mobile network cellular antennas, and demand of the users. While in-site measurements
networks and the operation of this type of system. This section are common, most of the study is generally performed with
presents a brief historic review of the generations until the forth. propagation models that can estimate the signal power received
Section 3 is focused in the fifth generation and its evolution. along the whole covered are. These models try to represent in
Some key technologies, mainly related to software, are presented a very compact way the complex propagation dynamics of the
in this section. In the Section 4, the Radio Access Network (RAN) wireless signal. Table 1 lists a small set of models available in the
is presented in more detail and some aspects of softwarization literature and their frequency ranges.
process is discussed. Section 5 describe the 5G core and most
of the several functions that compose the new architecture. In the Model Frequency range
Section 6, some aspects of the integration of a 5G system and non- Okumura-Hata 150–1500 MHz
3GPP access networks are presented. This section also discuss in COST 231 – Walfish-Ikegami 800–2000 MHz
more detail the integration of a 5G core and LoRa access networks. ITU-R P.529 700–3500 MHz
Finally, Section 7 summarizes the contributions expected in the TABLE 1
Examples of propagation models.
Releases 16 and 17, mainly in the software context.

2 M OBILE CELLULAR NETWORKS UNTIL THE Traditionally, a mobile cellular network operates in licensed
bands which must be acquired according to the specific set of
FORTH GENERATION
rules of each country and may be very expensive. Therefore, a
While the word ‘cellular’ has become almost redundant nowa- lot of the effort is devoted to managing the frequencies available
days when referring to mobile networks, it summarizes several for communication properly. This management involves several
concepts and innovations. A mobile radio (telephone) service was coordinated approaches to seek the efficient usage of the Radio
initially provided by high-power transmitters/receivers that could Frequency (RF) spectrum, which include antenna schemes, power
communicate directly to each other until a radius of about 80 km. control, interference control, frequency reuse, among others.
These devices were heavy, expensive, consumed a lot of energy, While all these concepts are still relevant in modern cellu-
and, as expected, had no support for data communication. The lar networks, new concepts and technologies such as Multiple-
introduction of the cellular radio changed the scenario drastically, Input Multiple-Output (MIMO), beamforming, and millimeter-
allowing the development of small, cheap, energy-efficient, and, wave communications have a significant impact. For example,
 3

Base station
transceiver
UE
R Base station
UE
controller Public
Mobile telec.
telecommunications
d Base station
switching office
switching
transceiver network
d d UE

UE Base station
controller

d d
Modern cellular
network Radio Access Network Core Network Data Networks
/ Internet
d

Fig. 3. Overview of a basic cellular system.

(a) Ideal network, offering interconnection between its UEs and the outside
networks. Several other management tasks are performed by an
MTSO to assist its UEs, such as channel (de)allocation, handoff,
paging, monitoring, and billing.
The channel (de)allocation may occur in different situations,
for example, when UE is turned on or off, when UE moves from
one BS coverage to another, if UE has a signal too weak, among
others. The handoff process consists of the coordinated transfer
of UE communication while moving between neighboring BSs,
involving signal monitoring and channel allocation and deallo-
cation. The paging process consists of finding a UE that has not
been communicating for some time and so its location is unknown.
Generally, the paging process is started by an incoming commu-
nication to UE. Moreover, the wireless signal quality, the cellular
(b) Practical (positions randomly generated in GNU
Octave) network, monitors the traffic of UEs to charge the users properly.
All these procedures remain in modern mobile networks, naturally,
Fig. 2. Representation of a cellular network. more focused on data instead of voice. Furthermore, technological
advances imply updating some concepts. For example, instead
of channel allocation and deallocation, modern mobile networks
beamforming makes it more complex to define cell coverage and may allocate and deallocate Physical Resource Blocks (PRBs) and
also to identify interference. Communications in the millimeter- millimeter-wave beams, and even edge computing resources.
wave bands (i.e., above 24 GHz) have very different propagation Similar to the traditional telecommunication systems, in a
characteristics than the traditional sub-6 GHz employed along cellular system, the only concerns of a user are to initiate or answer
decades in the mobile-cellular networks. a communication. The network system transparently manages ev-
erything, i.e., the user does not need to worry about configuration,
2.2 Basic operation of a cellular system reconfiguration, tuning, or any related task. Similar to traditional
In its original configuration, a cellular system could be sum- telecommunication systems, a cellular system has two types of
marized into the following elements: Mobile Telecommunications channels between UEs and BSs: control and traffic channels.
Switching Office (MTSO), BS, and mobile unit or User Equipment Control channels are used to exchange information related to
(UE). Fig. 3 illustrates a basic cellular system. In a modern cellular control and management tasks (e.g., UE initialization, handoff,
system, the MTSO can be identified as the core (network), while paging). Traffic channels are used to transport (voice or data)
the collection of BSs is named as Radio Access Network (RAN). communications. In modern cellular networks, a more generic
Actually, several other elements are composing the core and RAN, nomenclature is adopted, control plan, and user plan, but the
as we describe later in this and the following sections. meaning is very similar.
The BS controller handles the communication process between
the UE and the rest of the network. Each BS may simultaneously
2.3 Evolution of the mobile cellular networks
serve multiple UEs under its coverage area. One MTSO serves
multiple BSs, being connected generally through wired links, but By employing the arguable separation in ‘generations’, the
wireless links are also common. Traditionally, all communications characteristics of the mobile cellular networks, until the fourth
between UEs are established through a BS, even when they are generation, can be loosely summarized by the information pre-
close to each other. The concepts, technologies, and standards for sented in Table 2. A detailed description of the history of mobile
direct communication (known as device-to-device or D2D [14]) cellular networks can be found in [15], [16], since our concerns
between UEs have already been introduced in cellular networks. are only the main characteristics of each generation and their
However, when its effective adoption begins, it remains uncertain. relationship with the softwarization process. Naturally, the fourth
An MTSO connects to the public telephone or telecommunications generation deserves special attention since it is the first one
 4

designed to have data and software as relevant elements and also Communications (GSM) had great success, and, nowadays, some
due to its present market share. open-source software projects implement enough GSM features to
Nowadays, 1G is the only one without commercial deploy- have a basic operational cellular network. For example, OpenBTS
ments in operation, while the other generations still exhibit a and YateBTS1 offer open-source software for the Base Transceiver
significant presence. For example, 2G, with the lowest participa- Station (BTS), while OsmocomBB2 implemented an open-source
tion, still has more than 20% of the market share. The substantial GSM baseband software.
investment in the legacy hardware and new opportunities, such
as the support for IoT communications, can partially explain the Third generation – 3G
long-living of 2G and 3G networks. Until the 3G, there was a The benefits introduced by the second generation expanded
considerable fragmentation in the radio access technologies, which the global deployment and the interest of the general public in
is illustrated by the several standards available. A large number of mobile networks. In the early 90s, the ITU requested proposals for
players and the trade-offs of the competing technologies are some radio transmission technologies as part of the International Mobile
characteristics of the maturity level of the cellular networks at that Telecommunications (IMT) 2000 program. This program planned
time and helps to explain the fragmentation scenario. By the end to have a universal global system, but after many discussions and
of 2G and the beginning of 3G, the 3GPP initiative is introduced, disputes about patents, a family of 3G standards was adopted.
having great success in 3G and 4G. The fourth-generation is the 3GPP, for the Universal Mobile Telecommunications System
first focused on data communication. It offers, by design, proper (UMTS), and 3GPP23 , for the CDMA2000, became the main
support for the IP (Internet Protocol) stack and notably higher data driving forces in the third generation’s standardization process.
rate in comparison to the previous generation. Release 99 is the initial set of 3G UMTS standards published by
3GPP and the last one numbered according to the standardization
First generation – 1G
year (or close since it was finished in 2000). The next is Release
The first commercially deployed cellular networks, now la- 4 that describes the UMTS all-IP Core Network.
beled as 1G, were an extension of the Public Switched Telephone In comparison with 2G, the third-generation cellular net-
Network (PSTN). Therefore, the focus was on analog voice works increased the data rates, improved the quality of voice
communication, without any concern about data communication. communication, improved the efficiency in the use of the RF
Later, the network infrastructure was adapted to transport data but spectrum, enhanced the integration with IP networks, and en-
achieving only very low rates. The software in this generation hanced the support for new services. However, 3G still maintains
was proprietary, low-level, and embedded in specific hardware. backward compatibility with several previous technologies, e.g.,
The network control was digital, but voice communication was circuit switch is always supported. Although incompatible variants
analog. Despite the several limitations, the first-generation cellular multiple have been adopted, CDMA was largely evolved in 3G and
networks were a great success, motivating the evolution and giving became the dominant technology for wireless communication in
rise to a high-valuable and strategic industry. that generation. The software in 3G systems become even more
relevant than in the previous generation since the infrastructure is
Second generation – 2G
more complex, and data communication becomes more relevant.
Compared with 1G, the second-generation cellular networks However, proprietary software and vendor-specific appliances pro-
provided higher-quality signals, higher data rates, and greater tected by endless patents increase the costs and delay innovation.
capacity. 2G introduced digital traffic channels, i.e., 2G systems On the other hand, in the first years of the millennium, mobile
support digital data natively, and voice is digitally encoded before cellular networks were already essential, so the evolution of these
transmitting. Since all information becomes digital, 2G systems systems became unstoppable.
started offering support for encrypting both user and control traf- Similar to 2G, 3G also may have its life extended due to
fic. The digital traffic also allows the adoption of error detection IoT and also due to the more recent and significant investments.
and correction techniques, which improves the quality of voice During the deployment of 3G, the telecommunication sector faced
communication. 2G systems also introduced new services, such as an economic world crisis and expensive auctions for obtaining
the Short Message Service (SMS), that was very successful. Time- the RF spectrum. UMTS was the most successful standard in 3G
Division Multiple Access (TDMA) and Code Division Multiple and, similar to GSM, has some open-source software, provided
Access (CDMA) were also introduced, improving efficiency in the by the same developer groups. OpenBTS-UMTS4 implements the
use of the RF spectrum. While bit rates offered by 2G networks basic functionalities according to the Release 99, using OpenBTS
are too slow for many modern applications, they are suited for a as a framework. The project Cellular Network Infrastructure5
wide range of IoT demands. This suited possibility has motivated from Osmocom offers a complete implementation that allows for
infrastructure operators around the world to extend the life of 2G deploying an operational 3G system.
systems and vendors to revisit some products to improve relevant
performance metrics for IoT, such as energy efficiency.
Fourth generation – 4G
The software in 2G systems already had an important role, be-
ing responsible for the critical control and management tasks, in- The first billion of mobile subscribers was passed in 2002,
cluding monitoring and billing. However, proprietary technologies the second billion in 2005, the third billion in 2007, the fourth
and specialized appliances, i.e., combined hardware and software, billion by the end of 2008, and the fifth billion in 2010 [17].
are dominant. On the other hand, the adoption of solutions based
1. OpenBTS: http://openbts.org/; YateBTS: https://yatebts.com/
on standards for open systems, such as the Telecommunications
2. https://osmocom.org/projects/baseband
Management Network (TMN) defined by ITU-T, may be seen 3. 3GPP2 is not an evolution of 3GPP, but a different standardization body.
as initial steps to a more advanced softwarization process in the 4. https://github.com/RangeNetworks/OpenBTS-UMTS
telecommunication area. In 2G systems, Global System for Mobile 5. https://projects.osmocom.org/projects/cellular-infrastructure/
 5

1G 2G 3G 4G
Deployment ≈ 1980 ≈ 1991 ≈ 1999 ≈ 2009
Main services Analog voice Digital voice, SMS Digital voice, data packets IP packets
Data rate 1.9 kbps 14.4 kbps 384 kbps 200 Mbps
AMPS, TACS, NMT, GSM, GPRS, EDGE, UMTS (W-CDMA),
RAN LTE, WiMAX
C-450, TMA, RTM PDC, IS-95, IS-136 CDMA2000, TD-SCDMA
Core PSTN (SS7) PSTN (SS7, ISDN) PSTN, ATM, IP IP network (EPC)
3GPP initial standard - - Release 99 Release 8
Global market share (by 2019)1 0% 23% 25% 52%
1
https://www.statista.com/statistics/740442/worldwide-share-of-mobile-telecommunication-technology/
TABLE 2
Characteristics of the generations of the cellular networks.

During that decade, the volume of data traffic increased notably, are similar in terms of both performance and some technologies.
until the global mobile data traffic surpassed voice by the end For example, both are based on the use of Orthogonal Frequency-
of 20096 . Therefore, the plans for evolving the mobile cellular Division Multiple Access (OFDMA) to support multiple access
networks into the fourth-generation needed to be aggressive to to network resources. However, the OFDMA approaches are not
catch up with the demand and data-oriented characteristic. In compatible, and there are other relevant differences, such as
2002, the work on IMT-Advanced started by seeking to define the the backward compatibility and core network. Along this last
vision and requirements for the next generation of mobile cellular decade, the differences favored the 3GPP LTE-A as the leading
networks. The “VAN diagram”, introduced by the ITU [18] and 4G solution. On the other hand, WiMAX had an important
replicated in Fig. 4, summarizes some initial ideas for the fourth- role of being a high-quality and aggressive competitor and so
generation. In addition to the very high data rates for both mobile motivated the 3GPP consortium to notably improve the initial
and fixed/nomadic users, the IMT-Advanced also identified several LTE. Nowadays, WiMAX is being employed in important niches,
other requirements, including the following: such as data communications and information sharing on the
airport surface, wireless broadband communications for Smart
• Employ an all-IP packet switched network.
Grid network applications, specialized private networks for Oil
• Dynamically share the network resources to support more
and Gas companies, and wireless backhaul in general.
simultaneous users per cell.
Table 3 summarizes the main content introduced in each 3GPP
• Support smooth handovers across heterogeneous networks.
release related to the fourth-generation7 . In Release 8, the Evolved
• Offer high quality of service for multimedia applications.
Packet System (EPS) is primarily defined, being composed of
Evolved Universal Terrestrial Access Network (E-UTRAN) and
System Architecture Evolution (SAE). Relevant technologies,
IMT-Advanced
Systems beyond IMT-2000 encompass such as OFDMA and Multiple Input Multiple Output (MIMO),
Mobility the capabilities of previous systems New capabilities
of systems beyond are also added to the standard in Release 8. E-UTRAN is known
IMT-2000 as Long Term Evolution (LTE) while SAE is generally referenced
as Evolved Packet Core (EPC). Release 9 introduced the complete
High New Dashed line
means that the integration of the Home eNodeB (HeNB), evolved the Self-
Enhanced mobile
IMT-2000
IMT-2000 access exact data rates Organizing Networks (SON) and the multimedia broadcast and
are not yet
determined multicast service (eMBMS), and also added new spectrum bands
Enhancement (e.g., 800 MHz and 1500 MHz) for LTE operation, and inter-
operability between LTE, WiMAX, and UMTS. LTE-Advanced,
New nomadic/local
Low area wireless access introduced in Release 10, can be considered as a toolbox of
features that can be flexibly implemented on top of LTE, including
1 10 100 1000 carrier aggregation (up to 100 MHz), MIMO evolution (up to 8x8
Peak useful data rate (Mbps)
in downlink and 4x4 in uplink), relay nodes, and enhanced Inter-
Cell Interference Coordination (eICIC).
Fig. 4. IMT-2000 capabilities and systems beyond IMT-2000. After accomplishing the requirements of IMT-Advanced in
Release 10, 3GPP continued the standardization of some enhance-
In 2009, only two candidate proposals for IMT-Advanced ments from the previous features and also of new functionalities
were submitted to ITU: 3GPP LTE-Advanced (LTE-A) and and services. For example, Release 11 introduced carrier aggrega-
IEEE 802.16m (Mobile WiMAX Release 2 or WirelessMAN- tion enhancements, further enhanced ICIC (FeICIC), further SON
Advanced). In the following years, several deployments of LTE enhancements, RAN Enhancements for diverse data applications,
and WiMAX started around the world. Initially, these deploy- Coordinated Multipoint (CoMP), advanced IP interconnection of
ments were not IMT-Advanced, i.e., they were not precisely 4G services, among others. A priority in Release 12 was the use
networks, but they could be easily upgraded from LTE to LTE-A of LTE technology for emergency and security services. Other
and from WiMAX 1.5 to WiMAX 2.0. LTE-A and WiMAX 2.0 notable features added in this release included small cells and
6. https://www.ericsson.com/en/press-releases/2010/3/ 7. https://www.cablefree.net/wirelesstechnology/4glte/
mobile-data-traffic-surpasses-voice overview-of-lte-3gpp-releases/
 6

Year Release Content (extremely summarized) communities and supporters. For example, OpenAirInterface9
2008 8 LTE is introduced and srsLTE10 offer open-source code for fully-functional UE,
2009 9 Enhancements to LTE eNodeB (the central element of a 4G RAN), and EPC (core) to
2011 10 LTE-Advanced the 3GPP Release 10 or posterior. Before finishing this section,
2012 11 Enhancements to LTE-Advanced in the following, we briefly describe the LTE-Advanced, i.e., the
2015 12 Further enhancements to LTE-Advanced leading representative of a 4G system.
2016 13 Matching the increasing throughput demand
2017 14 First steps into 5G standardization LTE-Advanced
TABLE 3
Main content introduced in the 4G-related 3GPP releases.
Fig. 5 illustrates the EPS with only E-UTRAN as the access
network, i.e., 2G and 3G access networks, are not represented
in this figure. E-UTRAN is concentrated on the evolved Node
network densification, device-to-device (D2D) communications, B (eNodeB) in which all radio functionality is collapsed. As a
Machine Type Communications (MTC), and WiFi integration into network, E-UTRAN is simply a mesh of eNodeBs connected to
mobile operator’s offerings. In Release 13, 3GPP continued to car- neighboring eNodeBs. However, as we describe in Subsection 4.1,
rier aggregation to large aggregate numbers of carriers in different in a softwarized RAN, the eNodeB can be decomposed in two
bands and identified Beamforming and MIMO as key technologies parts (Baseband Unit – BBU – and Remote Radio Head – RRH),
to address the future capacity demand. Release 13 also included: bringing potential benefits such as resource pooling and energy
LTE in unlicensed spectrum (also known as Licensed-Assisted efficiency.
Access), enhancements for MTC, enhancements for D2D, and Traditionally, any cellular network may experience reduced
indoor positioning. data rates near the edge of its cells due to lower signal levels and
higher interference levels. In this context, an optional (but useful)
The Release 14 introduced several enhancements and new
element of the RAN is the Relay Node (RN) that has a reduced
technologies, including improvements in the Mission Critical
radius of operation compared to an eNodeB. Additionally, an RN
aspects (introduction of video and data services), the introduction
is more straightforward than an eNodeB, and so it can be more
of Vehicle-to-Everything (V2X) aspects, advancements in the
efficient in this context, working as an intermediary between the
Cellular Internet of Things (CIoT) aspects, improvements in the
eNodeB and UEs. On the other hand, an RN is not simply a signal
radio interface (in particular by enhancing the coordination with
repeater, since it receives, demodulates, decodes the data, applies
WLAN and unlicensed spectrum), enhancements for TV service,
error correction, and transmits a new signal.
multimedia priority service modifications, eMBMS enhancements,
among others. On the other hand, the discussion about 5G had a lot
User Plan (UP)
of attention in this release, and of the novelties in Release 14 were
Control Plan (CP)
introduced to contribute to the transition to the new generation.
For example, Control and User Plane Separation (CUPS) and en-
UE
hanced DECOR (eDECOR) are considered key features that help
to meet the way towards 5G8 . We present additional information eNodeB UE
about these two features in the next section, i.e., in the context of RAN -
UE RN E-UTRAN
the fifth-generation. UE (except UEs)
In terms of the software perspective, 4G can be considered
eNodeB
a noticeable evolution compared to the previous generations.
Several incentives (both positive and negative) have been driving
the softwarization process, including open and easily accessible
standards, the contribution of multiple communities in the
standardization process, pressure for lowing costs, pressure for HSS MME S-GW
facilitating innovation, pressure for multi-vendor integration, Core -
among others. On the other hand, some of the most powerful EPC
software paradigms, such as virtualization, software-defined PCRF P-GW

networking, microservices and, other cloud computing-related

ones, are still incipient or absent in the 4G systems. While several
basic functionalities of RAN and core can be implemented in External Services
Networks
general propose hardware, there are still many advanced features
that are only available with the help of specialized hardware
(e.g., Field-Programmable Gate Array – FPGA) and protected by
patents and copyrights (e.g., Intellectual Property – IP). Actually, Fig. 5. System architecture of EPS with E-UTRAN only.
patents and copyrights are not the issues by themselves, but the
vendor lock-in and barriers to innovation generally derived from A fundamental change in the architecture of the 4G core
them. Over the last decade, several open-source projects related to networks is the absence of circuit switching, i.e., the EPC does not
4G have been initiated. Many of these projects were discontinued, have direct connectivity to traditional circuit-switched networks,
mainly those related to WiMAX, but there are projects with active such as Integrated Services Digital Network (ISDN) or PSTN. As

9. https://www.openairinterface.org/
8. https://www.3gpp.org/news-events/1822-sa-rel-14 10. https://github.com/srsLTE/srsLTE
 7

a consequence, voice communication is transported over IP or, this significant and widely publicized evolution hides the real 5G
more recently, over LTE until the core. The main components of revolution, which we describe below.
the EPC are the following: 5G is being developed for the innovation and evolution of
cellular systems. This generation is the most dynamic and flexible
• Mobility Management Entity (MME) – This is the main mobile networks ever designed, making extensive use of native
control element in the EPC and operates only in the cloud applications and cores. Moreover, 5G can shaped during its
Control Plan (CP). The MME is involved in authenti- development to absorb new evolutionary leaps within the same
cation, security, mobility management (tracking, paging, pattern. This characteristic aims to accompany the exponential
handover), management of subscription profile, and ser- evolution of the leading technologies, such as artificial intelligence
vice connectivity (bearer setup). Since the first UE register and automation. To achieve this maturity, 5G was designed to
to the network, MME is involved in the authentication make extensive use of the virtualization of services and microser-
process that can be repeated later, including periodically. vices. At first, 5G will be operated by a hybrid network, in which
MME is also responsible for security measures such as convergence, integration, and coexistence with legacy systems
ciphering, generation of integrity protection keys, and are inevitable. However, in the future, the 5G system will be
allocation of temporary identity to the UE. Moreover, to predominantly composed of virtual network functions.
store and update UE-related data, mobility management The 3GPP designed the 5G radio access network called Next
involves the MME participating in the control signaling Generation Radio Access Network (NG-RAN) and the core with
with eNodeBs and S-GWs, eventually, with other MMEs. a Service-Based Architecture (SBA) independently and interop-
During the time that the MME is serving a UE, it stores erable. Based on this project, it is a different approach from
a temporary copy of the subscriber profile associated with previous generations of mobile networks, in which RAN and core
UE. This profile contains the information that MME needs were designed in a coupled manner. Therefore, as of 5G, the
to adequately assist UE, for example, to set up the bearers standard provides for the integration of elements from different
to transport the services requested by UE. generations in different configurations. The initial definition phase
• Home Subscriber Server (HSS) – This is the data repos- of the 5G architecture, i.e., the transition between 4G and 5G, a
itory for all permanent user data, such as a master copy terminology was defined to identify modifications that emerged
of the subscriber profile and the permanent key used to from the integration between RAN and core. The set options are
calculate the authentication vectors. Some temporary data illustrated in Fig. 6.
may be stored in HSS, such as the location of the user in
the level of the visited network (i.e., when UE changes of
EPC 5GC
MME) and identities of those P-GWs in use if the support
for mobility between non-3GPP access networks is active. 1
5
3 7 8 4
6
2

• Policy and Charging Rules Function (PCRF) – This net-

work element is responsible for Policy and Charging
Control (PCC). Basically, PCRF makes decisions on how LTE LTE NR NR

to handle the services in terms of QoS and provides the NR LTE

necessary information to the P-GW and, if applicable, also Core

RAN
LTE only NR only
LTE with NR for
data only
NR with LTE for
data only
to the S-GW, so that appropriate bearers and policing can EPC Option 1 (4G)
Option 6
(disregarded)
Option 3
Option 8
(disregarded)
be set up. 5GC Option 5 Option 2 Option 7 Option 4

• Packet Data Network (PDN) Gateway (P-GW) – This is

the interconnection point between EPC and external net- Fig. 6. Options defined during the initial phase of the 5G system defini-
works (e.g., the Internet). P-GW performs traffic shaping tion.
and filtering functions according to policies defined for
each UE and service. P-GW also collects and reports the In Fig. 6, option 1 corresponds to the existing 4G system
related charging information and, typically, allocates the and specified in previous releases of 3GPP. Options 6 and 8
IP address to UEs. were disregarded, as they would introduce many limitations to
• Serving Gateway (S-GW) – This is the interconnection the 5G New Radio (NR) to provide backward compatibility with
point between RAN and EPC. The main functions of S- EPC. Although options 2, 3, 4, 5, and 7 were considered in the
GW are User Plan (UP) tunnel management and switching. specification work, it was defined that the priorities would be
S-GW is minimally involved in control functions, being options 2 and 3. Therefore, Release 15 of 3GPP introduced the
responsible for its resources that are allocated based on Non-Stand Alone (NSA) architecture for option 3, and Stand-
requests from MME, P-GW, or PCRF. Alone (SA) architecture with the option 2.
The NSA architecture was proposed in May 2018, and six
months later, 3GPP presented the specifications for the SA ar-
3 5G NETWORKS AND BEYOND chitecture. NSA is a fast way to provide high data throughput
The evolution of mobile cellular networks has always sought and high connectivity, as it allows the use of existing network
to reach a ubiquitous high capacity radio, i.e., with increasingly assets, without the need to implement a new complete end-to-end
lower levels of latency and higher transmission capacity. At first, system for the 5G network, i.e., a transition architecture from 4G
the arrival of 5G adds to this evolution the opening of a path for to 5G. At the NSA, only radio technology needs to be updated.
IoT and intelligent communication technologies, exploring new However, to better exploit 5G’s capacity, aiming at new services,
forms of connectivity between Vehicles-to-Infrastructure (V2I), a modern architecture independent of the existing 4G system is
vehicle-to-vehicle (V2V), and Device-to-Device (D2D). However, necessary. SA is considered the ultimate 5G architecture, including
 8

improvements in radio transmission and a native 5G cloud core. represented by the blue circle) and eDECOR (with details in UE,
The two architectures proposed in 3GPP Release 15 are illustrated represented by the red circles) to access different DCNs.
in Fig. 7 and are discussed in the following subsections. EDECOR can be considered a precursor to the network slicing
concept, already introducing some characteristics similar to the
differentiation of QoS and the instantiation of members of the
3.1 5G Non-Stand Alone (NSA)
core network to serve each service. On the other hand, network
The NSA architecture, also known as EN-DC (E-UTRAN- slicing is a more advanced concept that allows the mobile network
NR Dual Connectivity), meets the telecommunications industry’s operator to perform end-to-end slicing, from a part of the radio
philosophy of introducing new technologies in the fastest and frequency, through the entire access network and including the
least disruptive manner. Although the motivations are clear, it is whole core. Network slicing should allow you to create resource
essential to highlight some disadvantages of NSA architecture. In slices that are partially or totally isolated, physically, or logically.
this architecture, NR can be implemented only where 4G/LTE cov-
erage already exists, indicating the absence of architecture NSA.
Furthermore, the available network functionality is limited to what Control and User Plane Separation (CUPS)
is offered by LTE/EPC. Therefore, several novelties introduced in As shown in Fig. 10, the control plane and the user plane of
the 5G system are not available, for instance, network slicing, each element (S-GW, P-GW, and TDF) use an Sx interface (a,
QoS treatment, flexibility in edge computing, and the general b and c, respectively) to carry out the necessary communication.
extensibility of the 5G core, including new applications in an The Sx interface offers procedures for establishing, modifying,
environment similar to the traditional cloud. and closing, to provide support for the control plan and user plan
The NSA architectural design focuses mainly on allowing at operations between the components of each separated node. 3GPP
least two 5G innovations to be introduced: (i) noticeable increase defined the Packet Forwarding Control Protocol (PFCP) to support
in bandwidth capacity and network throughput, and (ii) greater these functionalities in Sx, and this protocol was also adopted in
flexibility in the functions of the user plan provided by the 5G systems.
gateways (S-GW and P-GW) of the EPC core. As illustrated Using S-GW and P-GW as examples, it is interesting to
in Fig. 8, a separation between the control plans and data in highlight some design features of CUPS. It was evident that not
the gateways, and a double connection through LTE and NR all scenarios would require separation between the plans (control
are two significant features introduced in the NSA architecture. and user) and that the most common deployment scenarios would
Three sets of technologies were introduced to implement these have to contemplate the coexistence of nodes with and without
characteristics: (i) DEdicated CORE networks (DECOR) and separation in a single network. Therefore, it was defined that the
enhanced DECOR, (ii) CUPS and (iii) NR as secondary Radio separation between the plans should not have any impact on other
Access Technology (RAT). In the following, these technologies components of the core, such as MME, PCRF, billing/pricing
are briefly described. system, and subscription management system. CUPS was also
designed not to impact any procedure or protocol involving UEs
and RAN elements. Consequently, other aspects of the network
DEdicated CORE networks (DECOR) e enhanced
are unaware of whether S-GW and P-GW with which they interact
DECOR)
have separate control and user plans.
A network operator is recognized by its Public Land Mobile
Network (PLMN) identifier, which initially corresponds to a
mobile network core. However, even in 4G, there was a significant NR as secondary access radio technology
demand to make this approach more flexible and allow a single The dual connectivity (DC) of the NSA architecture allows UE
operator to instantiate multiple cores and direct each user to and RAN to receive and transmit data through two base stations
the appropriate core, according to the required service. Before simultaneously. In this way, a DC provides the ability to use radio
introducing eDECOR, the separation of cores was only possible resources used by two groups of cells operating but connected
using different PLMN identifiers, instantiating independent core to a single core. In the NSA context, the original motivation for
networks. An alternative solution was to use different Access Point dual connectivity is to increase user throughput. Furthermore, it is
Names (APNs) to direct users to various service networks, which also possible to provide greater mobility robustness and support
were associated with multiple user plan entities, in particular, P- load balancing between RAN nodes. Initially, the concept of dual
GWs. Both approaches, illustrated in Fig. 9(a), are not flexible. connectivity was introduced for EPS with two groups of cells,
Through the introduction of eDECOR, an operator can deploy giving E-UTRAN capabilities. BS connected by UE for first is
multiple Dedicated Core Networks (DCNs) within a PLMN with known as the master node and has the location where UE is
each DCN consisting of one or various nodes of the core network associated (with all signaling towards the core is carried out).
(e.g., only MME, MME with GWs, MME, GWs, and PCRF). Each When UE reaches the connected state, the master node can request
DCN can be dedicated to serving a different type of UE, separating another BS (i.e., secondary node) to offload data traffic. Later, the
certain types of traffic on specific core network nodes and, if solution evolved to support multiple radio technologies (MR-DC -
necessary, adjusting them differently from the rest of the nodes Multi-Radio DC), with EN-DC being the alternative that combines
of the core network. Including DECOR, the information needed the technologies E-UTRAN and NR.
to identify how to route user traffic is obtained by MME only in The requirement for spectrum flexibility was a determining
HSS. At the same time, eDECOR requires UE to provide specific factor for the adoption of technologies applied in OFDM in
information (i.e., the desired DCN) to facilitate fast and optimal LTE for 4G. It remains an important factor for planning and
selection of the dedicated core network. Fig. 9(b) illustrates the deploying NR to 5G. Throughout 3G and 4G, the need emerged
use of DECOR (with information initially only in the HSS, for allocations in different spectrum frequency bands, various
 9

Data
Data
SBA Control
EPC-CUPS network

P-GW-C P-GW-U NSSF NEF NRF UDM PCF NWDAF AF

HSS

CORE
PCRF S-GW-C S-GW-U UPF SMF AMF AUSP N3IWF

MME

CU CU

DU DU

RAN
RRH/RU RRH/RU
RRH/RU RRH/RU

NG-RAN NG-RAN

5G NSA architecture 5G SA architecture

Fig. 7. 5G architectures: Non-Stand-Alone (left) and Stand-Alone (right).

In addition to being a spectral band with a low occupancy rate

EPC
HSS PCRF and less crowded, the millimeter-wave band can allow communi-
S6a Gx cations with rates on the order of Gigabits per second, and it is
S11 S5-C
MME S-GW-C P-GW-C quite attractive for mobile networks. However, these waves cause
Sxa Sxb
significant challenges in the design, implementation, and operation
Data
of these systems, as they suffer more considerable degradation and
S-GW-U P-GW-U
S5-U network attenuation in the propagation. Moreover, these waves are more
susceptible to physical blockages and atmospheric absorption. On
S1-MME
S1-U S1-U
the other hand, with the development of increasingly smaller an-
X2-C tenna elements, thanks to its shorter wavelengths, the construction
LTE X2-U of antenna sets for the use of massive MIMO and beamforming
NR
(Master node) (Secondary node) is enabled, which helps in the treatment of the propagation
UE
problems and significantly improves the frequency reuse and the
spectral efficiency. However, beam formation/maintenance takes a
considerable amount of time, and its improvement has been the
Fig. 8. NSA architecture, with emphasis on the selection between control subject of several studies [21], [22]. Additionally, the problem
and data in GWs and double connection (LTE and NR).
becomes particularly challenging when considering the mobility
of users. One of 3GPP’s strategies to minimize these limitations is
operating bandwidths, several duplexation schemes, and multiple- to use the SA 5G architecture, as presented in the next subsection.
access schemes. However, the width and diversity of the spectrum
used in NR represent one of the most peculiar characteristics of 3.2 5G Stand Alone (SA)
5G. NR supports a large and diverse spectrum from 410 MHz to In the SA architecture, EPC is replaced by SBA (described in
52.6 GHz (and up to 100 GHz in future releases). Moreover, NR detail in Section 5), allowing the use of a set of virtual network
can use large bandwidths from 5 MHz to 3.2 GHz, supporting functions that help other functions in the architecture or even for
Time Division Duplex (TDD) and Frequency Division Duplex final users applications. In this way, an SA architecture consol-
(FDD), as well as additional carriers for descent, Supplementary idates the concept of decoupling of the data and control plans,
Downlink (SDL), or uphill, Supplementary Uplink (SOL). In allowing flexible and stateless positioning of virtual environments
this context, 3GPP defined at Release 15 two frequency bands in the different network segments that make up a 5G system. For
Frequency Range (FR): (i) FR1 (410 MHz - 7125 MHz) and example, these segments can be seen as Edge, Fog, and Cloud, or
(ii) FR2 (24250 MHz - 52600 MHz). FR2, also known as the even specifically in the RAN, we can find organized like Fronthaul,
millimeter-wave band, is the one that offers the greatest capacity Midhaul, and Backhaul.
but has different physical characteristics from FR1, presenting the A significant breakthrough in SA architecture is its direction
greatest challenge for deployment and use in 5G. In the following, towards the development of virtual network functions based on the
we present some additional information about millimeter-waves, cloud’s native model. This approach refers to how the functions of
which are still a subject of significant academic interest [19], [20]. virtual networks are created and deployed flexibly, widely using
 10

APN1=MBB modularity, reuse of functions, and interoperability with hetero-

GWs GWs geneous networks. Furthermore, the introduction of microservices
PLMN ID 1 HSS PLMN ID 2 and interfaces to access them simplifies the process of updating
CN 1 CN 2 and maintaining 5G systems, reducing the cost of operation, and
accelerating the introduction of new services.
MME MME GWs
APN2=CIoT
The full implementation of SA architecture presents some
challenges. For example, historically, the mobile network genera-
tions are strongly affected in the context of UE sessions, featuring
a state-based architecture. On the other hand, systems based on
the native cloud computing model have a fundamental feature,
stateless architecture, where the contexts of their states are saved
APN2=CIoT in the databases. For instance, the SBA core has introduced an
UE
UE
IoT dev unstructured data storage function (UDFS) to control the setting
APN1=MBB
of network functions, as can be seen in Fig. 11. This characteristic
allows the system to enjoy benefits such as elasticity and dynamic
(a) pre-DECOR
management of microservices on mobile networks. One of the
great challenges of 5G is to design and develop software com-
GWs GWs
ponents originally based in states, through the stack of protocols
HSS used by 3GPP, for stateless microservices, enjoying the benefits of
CN1 CN2
using the native cloud computing.
MME MME

Context
Network
UDSF data
function (state)
Network Context
function
data
Network
UDSF
function
Registration Session
...
Authentication control
UE IoT dev
eDECOR
DECOR

(b) DECOR e eDECOR Fig. 11. Context data function of SA architecture.

Fig. 9. Example of a pre-DECOR configuration, using APNs, compared In the SA architecture, the fundamental characteristic is that
to using DECOR and eDECOR.
the elements are defined as network functions, which provide
services to other functions, or any other authorized consumer,
EPC through programming interfaces. In this way, the architecture
Gx
PCRF
provides extensive modularization and reusability in complete
S5/8-C
isolation between DP and CP. Due to this high modularization,
S11-C
S-GW-C P-GW-C TDF-C the eventual migration of the NSA architecture to SA must be
Sxa Sxb Sxc hidden to the end-user. Moreover, it is worth mentioning that
NSA and SA are not competing for each other, but rather an
MME S-GW-U P-GW-U TDF-U
S11-U S5/8-U SGi evolutionary path for adopting the innovations introduced by 5G.
S1-U The intention is to start with the NSA and gradually migrate
to SA over time, especially for operators that already have
RAN
significant investment in 4G networks. For some time, NSA and
SA must coexist, and several approaches are possible. Fig. 12
illustrates a potential solution proposed by Samsung [23], in
which an intermediate architecture emerges, using a common core.
Fig. 10. Basic EPC architecture with the separation between the control
and user plans. Open-source software initiatives in 5G
The OpenAirInterface Software Alliance also has initiated the
development of software for 5G11 involving both RAN and
the concept of cloud computing to develop, deploy, and manage
core. Nowadays, only basic features are available [24], but it
services. For example, within this approach, the concept of mi-
is possible to observe the evolution of the new gNB and the
croservices can be used for the implementation and expansion
5G-NR. An operational 5G core from the OpenAirInterface is
of a 5G system, in line with the fast evolutions of information
still not available, but there are efforts in several components,
technology. One of the objectives of using microservices is to be
such as AMF, SMF, UDM, AUSF, and UPF12 . Another initiative
able to decompose the components into functions used, with low
granularity, to make the service light and with a high capacity for 11. https://gitlab.eurecom.fr/oai/openairinterface5g
sharing. This objective perfectly fits the need to define eMBB, 12. https://www.openairinterface.org/docs/workshop/8
mMTC, and URLLC communication scenarios since it offers Fall2019Workshop-Beijing/Training/2019-12-03-NGUYEN-DU.pdf
 11

is the free5GC, an open-source project focused in implementing the communication of the devices through various strategies such
the 5G core network (5GC) defined in 3GPP Release 15 (R15) as caching, downloading of processing, use of context information,
and beyond. The code is organized in stages, in which stage 1 allocation of bandwidth, among others. However, a reliable high-
is actually an EPC implementation improved with AMF, SMF band offer using millimeter-waves (later from the THz band) in
and UPF. Stages 2 and 3 are ‘pure’ 5G core implementations, in mobile scenarios is a challenge that will hardly be overcome on
which the latter is the most functional one. Yet another initiative a large scale in 5G networks. Due to the high directionality of
is the Open Core Network group13 that still does not have any the millimeter-wave beams, the alignment between the antennas
software available, but that is supported by the Telecom Infra of mobile devices and BSs must be remade with high regularity.
Project (TIP). TIP has been involved in several other open-source Currently, this means disruptions in communication, but as the
initiatives related to the telecommunications industry. number of BSs (both millimeter-waves and sub-6GHz) grows, it is
possible to mitigate the problem through the appropriate resources
NSA architecture SA architecture allocation policies. However, this allocation is challenging, as it
UDSF PCF involves multiple devices, multiple BSs, different resources, and
HSS PCRF must be performed dynamically in short time scales to meet the
NRF NSSF NEF
mobility of users [28]. Based on this context, there is a large
MME S-GW-C P-GW-C + SMF
amount of data that needs to be processed quickly for decision-
AMF making regarding resource allocation.
S-GW-U P-GW-U + UPF In addition to optimization models, the context described can
also make extensive use of machine learning and artificial intelli-
vCU gence techniques. These techniques were initially explored at 4G
Common through the concept of Self-Organizing Networks (SON) [29]. In
core 5G networks, machine learning and artificial intelligence should
have broader use, given the large volume of data to be manipu-
LTE DU NR DU lated and the greater complexity of the network [25]. However,
widespread adoption is expected only after the consolidation of
UE UE UE
5G networks and in the next generation of mobile communica-
tions [30]. Moreover, the increase in the volume of data and the
complexity of the infrastructure and recent evolution in machine
Fig. 12. Proposal for the coexistence of NSA and SA architectures.
learning and intelligence techniques have motivated its more
extensive use. However, for artificial intelligence-based solutions
3.3 Post-5G to be widely used in communication networks, they must be
Although many countries, such as Brazil, are still discussing built from the perspective of Explainable Artificial Intelligence
and preparing for the deployment of 5G networks, 2019 is the (XIA) [31]. Understanding the behavior of solutions based on
year when effective adoption began in several countries. Germany, artificial intelligence is an essential requirement for its adoption
China, South Korea, United States, Finland, and the United King- in communication infrastructures.
dom are examples of countries that are already deploying, testing, 5G networks are introducing new communication technologies
and effectively using 5G networks. As expected, the scientific to serve a wide range of applications and services. For example,
community has already begun to investigate questions about eMBB, URLLC, and mMTC are part of the 5G system to support
the limits of 5G networks, about applications (new or existing) applications such as 4K and 8K videos, autonomous cars, virtual
that would benefit from wireless communication. However, these and augmented reality, and Industry 4.0. However, applications are
networks would not be appropriately served about the technologies already emerging that are not adequately served by 5G networks,
that need to evolve or be created to meet new potential demands, such as multisensory extended reality, holographic telepresence,
among several other themes [25], [26]. Although it is not possible wireless brain-computer interactions, and connected autonomous
to predict what the next generation of mobile wireless networks and robotic systems [25], [26]. In general, these applications
will look like and which applications will justify their adoption, have new QoS requirements that can motivate the introduction
some issues are recurrent in the initial discussions and investiga- of new classes of service such as Mobile Broadband Reliable Low
tions and are likely to be part of the evolution of 5G networks. Latency Communication (MBRLLC) and Massive Ultra-Reliable
Among these themes are related with the efficient allocation of Low Latency Communications (mURLLC - massive URLLC). On
diverse resources in wireless networks (e.g., sub-6GHz bands, the other hand, to exploit the available resources more efficiently,
millimeter waves and, in the future, bands in THz [27]), wide it is vital to investigate and adequately characterize the demand for
use of machine learning and artificial intelligence, and applications applications, taking into account QoS requirements, Quality of Ex-
that are not served by 5G networks, such as multisensory extended perience (QoE), and quality of Physical Experience (QoPE) [25],
reality and wireless brain-computer interactions. [32].
5G networks have made the need to consider multiple types
of wireless resources notable, in addition to the traditional li- 4 T HE R ADIO ACCESS N ETWORK
censed sub-3GHz bands. In 5G networks, it is possible to use
the 3.5GHz band, unlicensed 5GHz band, millimeter-waves, and A RAN is responsible for managing the air interface to keep
D2D communications. Additionally, there are already Multi- a large number of users connected. Therefore, RAN needs to
access Edge Computing (MEC) resources that allow improving efficiently manage the RF spectrum, performing a series of tasks,
such as radio resource management, connection mobility control,
13. https://telecominfraproject.com/open-core-network/ dynamic allocation of resources to users’ equipment, compression
 12

Antenna RRH
RF

UE Coaxial UE
cable
S1/X2 BS Antena CPRI RRH
Interface Interface
RF

MAC

PHY

RF
UE
BBU
Coaxial UE
BBU

MAC

PHY
cable
Antena BBU RRH CPRI

MAC
BS

PHY
Interface

MAC
RF

PHY
MAC

PHY

RF
UE Coaxial S1/X2 UE
S1/X2 cable Interface CPRI
Interface Interface
BS
MAC

PHY

S1/X2 RF
Interface

2G RAN architecture 3G/4G RAN architecture

Fig. 13. Evolution of RAN architectures.

and security in the physical layer, session management, and QoS architecture, BBUs can be located in a data center (called cen-
flow, among other functions. tralized baseband architectures), allowing cost reduction through
Initially, on a traditional RAN, the RF front-end and baseband centralized maintenance and by the elasticity of cloud computing.
processing functionalities were integrated within a BS. The an- In this new configuration, RRHs can be geographically separated
tenna module was located a few meters from the radio module, from a set of BBUs by up to approximately 40 km. However, it is
connected with a coaxial cable, and showed high transmission important to note that a BBU is limited to processing RRH signals
losses. This architecture was noted on 1G and 2G mobile net- within a maximum distance, determined according to the delay
works. In recent years, the RF front-end has been decoupled from restrictions. This delay is mainly influenced by three factors: (i)
the baseband processing module, allowing mobile operators to distance between BBU and RRH, (ii) channel conditions, and (iii)
replace this module according to user demand. This architectural available processing capacity. According to Marotta et al. [33], the
decoupling allowed advances in centralization and virtualization processing capacity should be increased significantly for RRHs
in the baseband processing module and the RF front-end. Next, that are experiencing low Signal-to-Noise Ratio (SNR) and long
Subsection 4.1 presents the evolution of the concept of central- distance from the BBU.
ization of baseband architecture, and Subsection 4.2 presents the The concepts involved in C-RAN have been drawn attention to
virtualization of RAN, which started in 4G and continues to evolve the initial implementation of 5G (Architecture NSA described in
in networks 5G. Subsection 3.1), mainly considering the benefits of cost reduction
and maintenance. However, BBUs are hardware-based platforms
using specialized digital signal processors. Including a long-term
4.1 RAN centralization objective, it is essential to replace BBUs based on specialized
The high data transmission rate between the digital baseband hardware with software using general-purpose hardware, i.e.,
processing domain and the analog domain of the RF front-end Virtual BBUs (vBBUs). In the same way, we can think of a
with the antenna requires a high bandwidth bus that connects these radio virtualization layer that allows several heterogeneous access
two domains. For many years, this requirement has limited the technologies coexisting on the same RRH. This coexistence uses
BS design to specialized hardware components with that high- innovative baseband processing techniques to divide and abstract
performance bus. This limitation was overcome with the use of an RRH into multiple virtual RRHs (vRRHs). This vision of
fiber optics on this bus, reducing data loss and increasing the RAN virtualization is aligned with the SA Architecture (described
distance between the connected domains. In 3G/4G generations, in Subsection 3.2) proposed in Release 15 of 3GPP, where, for
especially since Release 8 of the 3GPP, the baseband processing example, the use of vBBUs is an Network Functions Virtualization
started to be implemented in BBUs, i.e., dedicated and specialized (NFV) use case to provide services on a 5G system.
hardware that implements a RAT. At the same time, RRH inte-
grates the RF front-end with the antenna. In Fig. 13, the evolution
4.2 RAN virtualization
of the RAN architecture from 2G to 3G/4G is illustrated.
In 3G/4G, the RF front-end and the baseband processing are The RAN virtualization (vRAN) has received prominence in
separated, as can be seen on the right side of Fig. 13. RRH also 5G systems because it allows us to create, manage, and configure
called a Radio Unit (RU), has a fiber optic interface and performs RANs dynamically, meeting specific requirements of each service.
analog/digital conversion and vice-versa, power amplification, and Furthermore, the vRAN concept opens up new business models
signal filtering. The baseband processing, now performed at BBU, in which service providers can rent vRANs from infrastructure
is isolated and independent from RU. This architecture is consid- providers. In this scenario, the infrastructure provider controls the
ered decentralized since BBU performs its operation separately entire physical resource, including the RF spectrum, the physical
from RU. Recent advances have boosted the bandwidth of optical RRH, the hardware resources in the data centers (i.e., servers with
fibers allowing the incorporation of cloud-based architectures processing, memory, and storage) and the physical network. A
into the Cloud Radio Access Network (C-RAN). In this type of service provider can hire from one infrastructure provider one
 13

or more vRANs, including at least one virtual BS, i.e., a vRRH 4.3 RAN demonstration
connected to a vBBU. Goals
Isolation, programmability, and adaptability are essential prop-
erties for customizing vRAN to accommodate the different ser- One objective of the demonstration is to present practically
vices provided for 5G networks. For example, the combination the functionalities of a RAN, using open software and hardware.
of BBU and RRH slicing can instantiate an end-to-end vRAN Another goal is to briefly comment on the software installation
over physical infrastructure. More specifically, slicing and virtu- and configuration processes that implement the RAN, making
alization can be implemented in a BBU to allow multiple vBBUs available material for replication of this demonstration.
to run on the same physical hardware. Likewise, one RRH or a
combination of them can support multiple vRRHs. These elements Description
applied to vRAN, constitute the pillars to provide multi-services In this demonstration, we create an operational eNodeB, i.e.,
for future mobile networks. the main element of a RAN, based on LTE technology and using
A set of vBBU and vRRHs can be run on General Purpose open-source software. In addition to the RAN, the software is
Processor (GPPs), taking advantage of the highly optimized signal also capable of emulating functional UEs. The demonstration
processing libraries and taking advantage of the ever-increasing is organized in 3 experiments: (1) UE and RAN emulated by
evolution of processors, such as higher processing power and software, without the core; (2) UE, RAN and EPC core, all im-
energy efficiency, such as can be seen in Fig. 14. Recently, the plemented in software; and (3) UE in hardware (conventional cell
3GPP RAN3 working group [34] considered dividing a vBBU into phone), RAN in hardware (SDR - Software-Defined Radio) and
two new entities, named Data Unit (DU) and Control Unit (CU). software, and EPC core implemented in software. All components
DU can host time-limited physical layer functions, while CU hosts are implemented using Docker containers that can be hosted on
non-critical function resources, such as the MAC layer and higher a cloud infrastructure. Fig. 16 shows the software and hardware
control services. DU implementation is expected to cover an area components that are used in the RAN demo.
of 10 to 20 km in radius, while CU implementation should cover
areas from 100 to 200 km.
In Release 15 of 3GPP, the cloud concept was maintained Additional information
in the RAN, but the name was changed to NG-RAN using an During the tutorial, demonstration videos of the experiments
interface called NG. Due to interoperability between 4G and 5G, are presented. Furthermore, manuals are available with details
DU and CU have been renamed gNB-DU and gNB-CU, respec- on how the tests can be replicated. Finally, the containers and
tively. A gNB is responsible for some tasks, such as radio resource any extra code produced by the authors needed to replicate the
management, connection mobility control, dynamic allocation of experiments are also publicly available.
resources to users’ equipment, physical layer compression and se- The repository of this tutorial:
curity, session management, and QoS flow, among other functions. https://github.com/LABORA-INF-UFG/NetSoft2020-Tutorial4.
Therefore, the gNB protocol stack is detailed, considering PHY,
MAC, RLC, PDCP, RRC, etc.
Virtualization also introduced the possibility of dividing the 5 C ORE NETWORK
gNB protocol stack into network functions. In this context, 3GPP Based on the mobile network scope, the core can be character-
proposed eight options for the functional division between central- ized as the most critical element in the 5G system. 3GPP Release
ized and distributed units [34], considering transport requirements, 15 [6] defined the core as a set of interconnected components
in particular flow and latency. González-Dı́a et al. [35] evaluated, by a service layer. Each component has specific responsibility for
in an experimental setting, three slicing options for fronthaul, in consuming and providing services for the other elements of the 5G
addition to an option for the DU backhaul, as illustrated in Fig. 15. system, through Application Programming Interface (API) defined
These eight slicing options proposed by 3GPP are still the subject by the standard. The structure composed of software components
of research [36] and development by academia and industry, as and their interconnections via APIs is shown in Fig. 17. Release
requirements for virtualization, processing, and functionality still 15 introduced substantial changes in the way mobile networks
need to be investigated. are designed to support a wide range of services, each as distinct
It is important to note that this new flexible architecture performance requirements [9].
of RAN composed of distributed data units (DU/gNB-DU) and Given the 5G core relevance, each of the elements that com-
control (CU/gNB-CU) brought changes to the transport network pose the core needs to be robust, resilient, higher availability,
between the access and the core of the 5G system. Currently, the and scalability. These characteristics are usually associated with
transport network is being redesigned and developed considering the cloud computing concept. In the Release 15 context, each of
segments of fronthaul, midhaul, and backhaul. The fronthaul the components that compose the new 5G core can be split into
is responsible for the communication between RRH/vRRH and Virtual Network Functions (VNFs). In this way, each VNF can
DU/gNB-DU. In the midhaul segment, communication takes place be available on-demand in a cloud infrastructure. The use based
between DU/gNB-DU and CU/gNB-CU. Finally, the backhaul on the split of VNF is possible because physical interfaces like in
performs the communication between CU/gNB-CU and the new previous architectures do not interconnect the components end-to-
core of the 5G system. Additionally, approaches are expected to end. Instead, each of the components of the core must expose a set
adopt the integration between transport segments, in a configura- of software functionalities through SBA [37]. In this context, each
tion of crosshaul [35], [36], in which the objective is to explore the VNF offers one or more services for other VNFs. Considering
efficient use of resources high-cost transportation. Orchestrating the new definition of the SBA model, the exposure of functions
the workloads of vRANs in this new architecture is a subject of applies only to the signaling context and not to the transfer of user
considerable research today. data.
 14

RRH
RF

UE
vBBU
vBBU CPRI
Interface
RF
vRRH vRRH
UE
Supervisor
CPRI
vBBU vBBU Interface
vRRH vRRH vBBU RF
vRRH
Supervisor UE
Supervisor CPRI
Interface

Cloud processing

Fig. 14. Virtualized RAN.

Fronthaul
5.1 Main Components

3GPP slices Bottom Top In this subsection, we introduce the components needed to
C-RAN Backhaul
slice slice create a basic 5G core. In this way, we have named these
RRC components of main.
L3 CU
PDCP
Access and Mobility Management Function (AMF)
RLC high
CU Mobility can be considered the essence of a 5G system. The
RLC low
L2 CU mobility management begins when a new connection is established
MAC high DU between a UE and the core of the network. This action triggers a
set of procedures to identify UE, providing a security structure
MAC low DU
to offer a transport channel of messages. The main goal of the
PHY high AMF component is to ensure that the communication process
L1 occurs cohesively and transparently, considering user mobility as
PHY low DU
a critical factor. Based on the functions implemented in AMF, the
Radio frequency DU network can reach a specific user to notify about any messages or
calls received, for example. Moreover, the AMF component can
allow a particular UE to initiate a communication process with
Fig. 15. Division of functions between the central and the distributed
units. other UEs connected to RAN or with Internet access. Another
fundamental functionality of AMF is to guarantee the connectivity
holding the existing sessions when UEs move between different
access points.
One of the communication methods used for an SBA is In 5G networks, there is a need to provide flexible support
based on Representational State Transfer (REST) over HyperText for a wide range of new users [9]. Many of these users have
Transfer Protocol (HTTP). This method consists of a set of rules specific requirements concerning mobility. For example, a par-
and guidelines, widely used in the interconnection of distributed ticular UE used in a factory does not usually move, while UE in
systems. The rules define how communication technologies on an autonomous vehicle or remotely controlled can present high
the Web access services through the use of APIs. The 3GPP mobility. To better support these different needs, 3GPP Release
expectation as this initiative is to make the task of extending the 15 [6] splits the mobility procedures into three categories for the
network resources simpler. Another aspect regarding the commu- AMF component:
nication mechanism is that it should be seen as a logical option
• Standard procedures can be characterized as a set of steps
since the components specification for VNFs assumes that these
running when any UE requests a connection to the core.
components are running in a virtualized environment [38].
Among these steps, the security process is highlighted,
The components of the 5G core can be seen as an intercon- composed of primary authentication, management of ac-
nected network of services. Each VNF of a component performs cess keys, identification, and basic configuration of UE.
specific responsibilities and interconnects with other VNFs pro- • Specific procedures have the function of managing the
ducing and consuming services. The main characteristics of each registration and periodic updating of the mobility of a
of these components are discussed in the following subsections. given UE in AMF. Moreover, this procedure controls the
We briefly describe the main components of the 5G core without closing of the UE registration, provide scope for different
the objective of listing all or making an exhaustive presentation. access technologies.
The complete list of components and their detailed description • Connection Management procedures are used to establish a
requires hundreds of pages of the standard or a whole book on secure communication process between a given UE and the
it [39]. core. Moreover, this procedure works when a specific UE
 15

Software Software Hardware + Software

UE
UE EPC UE EPC

Phone SDR
(1) (2) (3)

Fig. 16. Experiments with a softwarized RAN.

as part of a split process between CP and DP, initially introduced

NSSF NWDAF AUSF UDM NEF NRF in the Release 14 [40] with CUPs. Decoupling data and control
allows the SBA model to decentralize its components further. For
APIs example, it can direct activities such as packet processing to be
(HTTP/JSON)
SMF AMF PCF AF
placed closer to the edge network, increasing QoS for the user,
and reducing network traffic.
SMF controls the functionalities implemented in the UPF
UPF N3IWF SBA component. The main function of UPF is associated with the
forwarding and the data processing from UEs. This component
is also responsible for generating notifications related to data
Fig. 17. Core 5G software components. traffic and the inspection of packets. UPF also works as a stable
interconnection point between the core and any external networks.
Moreover, UPF allows communication to happen transparently,
needs to carry out a network resource reservation process hiding complex aspects associated with mobility. IP packets des-
for sending data. tined for a give UE are forwarded (from the Internet) to the
Each of these categories of procedures aims to provide UEs respective UPF, which is attending that UE, even when UE is
with functionalities to establish connections with the core network, on the move. In general, the UPF component is responsible for:
using the services associated with mobility. • Design the rules for the interconnection point between the
core and any external networks;
Session Management Function (SMF) • Serve as an external access point for Protocol Data Unit
The SMF component is responsible for managing the UE ses- (PDU), interconnecting different data networks;
sions, i.e., the sessions representing the connected users. The main • Forward and route the packets and inspect these packets
responsibilities of SMF are activities of establishing, modifying, aiming to detect application characteristics;
and releasing the individual sessions of UEs and allocating the IP • Apply the definitions associated with the management of
addresses for each connected UE. However, the communication the user data plane and provide information on the data
between UEs and SMF is performed indirectly through the AMF traffic.
component. The function of AMF forwards the messages associ-
ated with the session of a given UE and the functions of the SMF Authentication Server Function (AUSF)
component. AUSF is responsible for the authentication service of UEs
The SMF component’s internal functions interact with VNFs through the access credentials provided by UDM. Moreover,
of the other components through the producer/consumer model, AUSF provides services associated with the encryption for allow-
defined regarding the SBA [37]. For example, SMF is responsible ing secure information traffic and the execution of the roaming
for controlling different functions associated with the UPF com- update processes, as well as other parameters related to UE.
ponent. This control includes the SMF’s ability to configure the Generally, the AUSF component provides services consumed
direction of data traffic associated with a UPF for a given UE by AMF functions, which request resources associated with the
session. Moreover, SMF must run monitoring and control actions authentication process. For example, AUSF internally processes
on UPF. The SMF component interacts with functions related to the requests of functions of the AMF component and after del-
PCF, aiming to execute the session policy of the UEs connected. egates to services provided by the UDM component to run the
This execution action can be described as one of the main tasks registry procedures in the data repository.
associated with a 5G system. For instance, this action determines
the guidelines for data connectivity between a UE and the Data Unified Data Management (UDM)
Network (DN).
The UDM component manages the users’ data on the network
in an only centralized element. UDM is similar to the HSS of
User Plane Function (UPF) the EPC core of the 4G networks. Through UDM, various VNFs
3GPP Release 15 [6] introduces the UPF component as a of the SBA model can run enumerates actions, such as registra-
fundamental function within the new SBA model. UPF can be seen tion and authentication of UEs, user identification, application
 16

of access rules and authorization, etc. The UDM component element, related to the PDU session in which AF is
interacts directly with AMF, which forwards the requests from associated;
other components. Moreover, in scenarios where exist more than • PCF-SM-PolicyControl – provides to SMF component
one instance of the AMF component on the network, UDM must guidelines for access related to PDU session in which the
control which instances are responsible for serving a give UE [41]. component is associated;
Among the UDM functionalities, we can highlight [42]: • PCF-BDT-PolicyControl – supplies a set of guidelines for
the NEF component to be used by applications to transfer
• Generation of the Authentication and Key Agreement;
data in the background plan;
• Handling user identification;
• PCF-UE-PolicyControl – provides control guidelines to be
• Support for privacy-protected signature identifier hiding;
used in the communication process management between
• Access authorization based on the signature data (e.g.,
UEs and other network functions;
restrictions associated with mobility)
• PCF-EventExposure – allows other network functions to
• Signature management;
register to be notified when a given event happens.
• SMS management.
The decision on the application of the monitoring and control
The UDM component works as a front-end for the user’s
policies made by PCF is based, in part, on analytical information
signature data recorded in the Unified Data Repository (UDR).
provided by other network functions, such as NWADF. Moreover,
UDM uses this signature data to execute the logic of several ap-
PCF is a fundamental component in a scenario where an AF needs
plications, such as access authorization, registration management,
to perform a given activity, i.e., data transfer in the background
and accessibility for finalizing events. UDR is a database where
plan. In this case, AF can contact PCF to infer the best time
various types of data are stored, and the access is offered as a
interval for the activity’s execution. This behavior allows the
service to other components such as UDM, PCF, and NEF. There
system operator to offer information to application providers about
is an optional storage component called Unstructured data storage
the most appropriate time to transfer data in the background plan.
function (UDSF), which allows other components or functions
to record dynamic context data outside the function or component
itself. In the 3GPP context, unstructured data refers to the structure Network Repository Function (NRF)
is not defined in the specifications, allowing each provider to use a The NRF component is the repository, where all the functions
give UDSF and choose its appropriate structure for storage. There available for a given network are listed. The goal of this compo-
is no requirement for any access compatibility or data storage in nent is to allow VNFs can find the appropriate function to meet
UDSF from different providers. their requirements. NRF has the responsibility to select the most
suitable service provider component based on the performance
Policy Control Function (PCF) criteria provided. In this way, the NRF component is updated
whenever a new NBF instance is deployed or modified. Moreover,
This component performs the same function as PCRF of the
NRF holds information on the other VNFs, such as type, capacity,
EPC core in the 4G system. PCF is responsible for controlling the
address, etc.
behavior of the network, applying security and control rules, re-
Based on the SBA scope, the NRF component plays a funda-
lated to session management, mainly for functionalities associated
mental role in the functioning of the other VNFs. This component
with user mobility. This component interacts directly with AMF,
offers a central mechanism that allows automating the configura-
providing an access and mobility policy that can add the control
tion process necessary for the other VNFs to discover and connect
of access restrictions to services in a given area, for example.
to specialized services.
Moreover, PCF can include the management of topics associated
with priority access to the channel of given UEs to the detriment
of others. This management is named Radio Frequency Selection 5.2 Additional Components
Priority (RFSP). In this subsection, we present additional components of the 5G
In the session management context, PCF can interact with the core. These components are fundamental for creating a complete
application functions and SMF. The main goal is to provide met- infrastructure of network functions for providing various business
rics associated with QoS and information regarding the data flow, opportunities.
which is obtained by regularly monitoring events related to the
PDU session. Moreover, PCF offers security policy information Network Slice Selection Function (NSSF)
for UEs. These policies can be associated with network resources
In the context of the 5G networks, Network Slicing can be
and rules for selecting resource slicing. For example, PCF can
defined as a possibility for the system operator to allocate a
be triggered to provide information when a given UE performs
resource set (in general, virtualized), to meet the requirements of
an access selection (UE access selection) or when a PDU session
given services or applications [43]. The aim is to offer support
is established [39]. The interaction between PCF and the other
for a wide range of services, each with specific performance
application functions is implemented through the exposure of six
demands [9]. The virtual resource slices are logical instances of
services, namely:
network resources needed to meet a given request. A slice can
• PCF-AM-PolicyControl – provides information on access include RAN and core resources, and even extrapolate to transport
control policies, network selection, mobility management, networks between various RANs and cores, considering network
and guidelines for the selection of routes between UEs and the end-to-end network service.
AMFs; The 5G core architecture defines the NSSF component as be-
• PCF-PolicyAuthorization – supplies authorization and pro- ing responsible for managing the available network slice instances.
vides access control policies to a request for an AF This component selects the network slice instances and the set of
 17

AMFs available for a given UE. AMF can be a component dedi- An essential factor that must be evaluated by the system
cated to a specific slice or to serve a set of network slice instances. operator is the confidence degree that an AF component can
The NSSF function is to assist AMF in choosing the available have to interact directly with specific VNFs. For example, an AF
network slices, redirecting traffic between the controlled network with higher reliability can access VNFs from all components of
slices. Moreover, NSSF can be seen as an orchestrator that can the SBA model, directly. At the same time, a lower reliable AF
influence how network traffic is routed. This component produces must first interact with the NEF component before accessing more
two services, a selection service that provides information about sensitive network functions.
the selected network slice and another availability service that
generates information concerning the available resource slices.
Non-3GPP InterWorking Function (N3IWF)
Network Exposure Function (NEF)
The N3IWF component is used to integrate non-3GPP ac-
The NEF component is responsible for exposing some internal cesses with the 5G core. WiFi (IEEE 802.11) and Data Over Cable
events related to UEs and the SBA model. The exhibition of these Service Interface Specification (DOCSIS) are examples of non-
events aims to meet the demand for specific applications and VNFs 3GPP access technologies intended for integration by the standard.
of other components. For example, these demands need to access The conventional 3GPP access uses a BS, e.g., eNB (4G) or gNB
to the location or notify about the connectivity interruption of a (5G). However, the non-3GPP access starts on a different device,
given UE. Moreover, this information allows the AMF component for example, a WiFi access point or a Hybrid fiber-coaxial (HFC)
to adjust the system according to user group behavior. modem. This device uses the N3IWF component to access the
The possibility of exposing internal events through a NEF 3GPP network and other 5G core components.
access interface opens new business opportunities for service All traffic from the N3IWF component is sent through secure
providers, allowing in some cases, more advanced services to be channels, and it is isolated from all 3GPP traffic. The isolation is
offered by third parties. For example, an application can use the maintained not only for data traffic (usual for 3GPP communica-
functions exposed by the NEF component to know if a given UE tions) but also for control traffic, including the traffic before the
is accessible or not. Moreover, these functions can determine the authentication process. More details about the N3IWF component,
geographical location of UE or if its device is in the movement. with its use and the interaction with other SBA components, are
The NEF component responds to requests from various VNFs described in Subsection 6.1.
through regular interactions between UDM and AMF components.

Network Data Analytics Function (NWDAF) 5.3 5G core demonstration

NWDAF is an optional component in the core architecture Goals
and is responsible for collecting several types of information from
the network and its users. This information is later organized and The main goal of this demonstration is to present in a practical
analyzed to provide the inferred results for other network func- way the main functionalities of the 5G core based on the SBA
tions. The description of this component was written superficially model. Moreover, we comment on the installation and configura-
in Release 15 [6], and its detailed specification is expected for tion processes of software that implements the 5G core, providing
Release 16. material to replicate this demonstration.
The data collected by NWDAF comes from several other
network functions that compose the core. The data collection is
made through the services layer that connects the core components Description
via the writing service. This service can be activated by the internal
events triggered by each component. Moreover, NWDAF collects In this demonstration, we use the 5G core based on the open-
information about the operation of the system and data registry source SBA model. We also present how to emulate RAN and
information in the UDR component. multiple UEs. The presentation is organized in two experiments:
Based on the specification, the services provided by NWDAF (1) UEs, RAN, and core, all implemented in software; (2) UE
can be consumed by any core component. The external access in hardware (conventional mobile phone), eNodeB in hardware
is also possible using the NEF component. The analyzes made by (SDR) and software, and the 5G core implemented in software.
NWDAF on the data collected over time can be used as a historical All components are implemented using Docker containers that can
and statistical resource to predict future values. Moreover, the be hosted on a cloud infrastructure. Fig. 18 shows the software and
analytical data produced by NWDAF can be used to apply specific hardware components that are used in the 5G core demonstration.
actions in the network context.

Additional information
Application Function (AF)
AF is a generic component that represents a possible applica- During the tutorial, we present demonstration videos of the
tion, internal or external, to the operator’s network, which interacts experiments. Moreover, the manuals are available with details on
with the SBA model. The interaction process of AFs with SBA can how the practices can be replicated. Finally, containers and any
influence some aspects of the whole system. For example, an AF extra source code produced by the team and need to replicate the
can interact with the PCF component through services exposed by experiments are publicly available.
the NEF component, improving QoS aspects and, consequently, Repository for this tutorial:
the charging and pricing policies. https://github.com/LABORA-INF-UFG/NetSoft2020-Tutorial4.
 18

Software Hardware + Software

SBA UE SBA

UEs Phone SDR

(1) (2)

Fig. 18. Experiments with a 5G core.

6 I NTEGRATION OF 5G SYSTEM WITH NON -3GPP Considering the focus of this article, no further information on the
ACCESS NETWORKS EPC core approach will be present, but a full description can be
found in [50]. For the 5G SBA core, fewer options were defined,
Although 5G networks have been designed to serve a wide
basically, untrusted non-3GPP access (Release 15, frozen in
variety of services, offering a board range of resources, 3GPP
March 2019 and completed in June 2019) and trust (Release
envisaged in Release 15 the integration and support of other non-
16, expected to freeze in March 2020 and conclusion to June
3GPP wireless access technologies, with particular attention to
2020). In this context, the term untrusted means that the 3GPP
WiFi (IEEE 802.11). As already discussed in some investiga-
network operator does not trust in the security offered by the
tions [44], [45], the coexistence of different access technologies
non-3GPP access network. Therefore, it needs to take actions that
promotes significant gains concerning the performance and cost
ensure the proper transport of traffic from this access network.
of the communication of IoT devices. Moreover, many devices
This need means that the non-3GPP traffic must be isolated
have a non-3GPP wireless interface, and these devices will be
from other traffics, including the 5G core, which is suitable for
in operation for some years. The integration of multiple wireless
IoT applications and services. The trust access does not allow
access technologies in an only core can benefit both users and
integration with other wireless communication technologies used
administrators of public and private communication infrastructure.
in IoT, such as LoRa or ZigBee.
Release 15 [46] already defines how non-3GPP wireless net-
work technology can be used as access and integrated into a To support the untrusted non-3GPP access network, the main
5G core. However, this release focus only on Wireless Local component introduced by 3GPP Release 15 was N3IWF, which
Area Networks (WLANs), i.e., WiFi networks, considering the is responsible for forwarding signaling and data between the 5G
untrusted access. The trusted access will be introduced in Release core and the non-3GPP access network, as described in Subsection
16 [47]. In this release, attention is also being taken to cable access 5. Fig. 19 illustrates the integration of two non-3GPP access
technologies, in particular for HFC. Although 3GPP standards do networks to a 5G core, showing the main components involved
not explicitly describe how to integrate other wireless communi- and their communication interfaces. N3IWF selects AMF to serve
cation technologies used in IoT, such as LoRa and ZigBee, the the IoT device (or gateway), which will be responsible for holding
framework allows extensions to be exploited [48], [49]. In the the (eventual) mobility of the equipment and brokering all the sig-
following, we present more details about Release 15 and also on naling with other 5G core functions. For effective communication,
LoRa/LoRaWAN technology, which we use as an illustration for a device needs PDU sessions, which are established, modified,
the integration between the 5G core and a non-3GPP wireless and released under the SMF component’s control. This component
access technology at the end of this section. works as a CP that operates on a DP implemented through the UPF
component.
6.1 Non-3GPP Access Networks
Non-3GPP
N1
The 5G core architecture foresees, since its design, the pos- networks
Core - SBA
sibility of integrating non-3GPP access technologies, with the WiFi
NWu
premise that the communication interfaces offer IP connectivity, UE
Y1
i.e., a traditional IP stack. Although this premise is met for AMF SMF
N11
technologies such as WiFi and HFC, several IoT solutions such as
N2 N4
LoRa, ZigBee, nRF24, and others consider that exists a gateway Y2
LoRa gw
to provide the IP connectivity for the IoT devices. Therefore, N3IWF UPF
IoT dev N3 N6 Internet
we assume from this point in the article that equipment with a Y2
traditional IP stack, i.e., a device or a gateway, is available to IoT dev NWu
Y1 N1
establish the integration with the non-3GPP network. It is essential
to highlight, as describe by [49] that there are several ways to
integrate devices that depend on a gateway with a traditional IP
Fig. 19. Integration between the untrusted non-3GGP access networks
stack. and the 5G SBA core.
The EPC core from 4G also has support for the non-3GPP
access network, using a more sophisticated approach. Moreover, In the 5G SBA core, the AUSF component allows a device to
for defining between trusted and untrusted access, it needs to authenticate itself to access the network and its services using
choose between network-based mobility and host-based mobility. a 3GPP wireless interface (e.g., NR) and a non-3GPP inter-
 19

face (e.g., WiFi). Non-3GPP devices can authenticate with the • Spreading Factor (SF) with values of 7, 8, 9, or 10. The
SBA core through a certification-based scheme, using Extensible higher SF, more information is transmitted per symbol,
Authentication Protocol-Transport Layer Security (EAP-TLS) or generating a gain in transmission;
EAP-Tunneled TLS (EAP-TTLS) and the traditional procedure • Bandwidth (BW) of 125 kHz, 250 kHz, or 500 kHz
with credentials based on Subscriber Identification Module (SIM). for a given SF. A narrower BW increases the reception
Fig. 20 shows the connections established to provide integration sensibility and increases the packet transmission time;
between the untrusted non-3GPP access network and the 5G SBA • Code Rate (CR) is responsible for detecting and correcting
core. errors.
These configurations determine the bit rate transmitted, the
maximum size of the transmitted data, and the transmission time
N1 of a packet in the RF spectrum. These configurations also influence
WiFi NWu
the size of messages, their range, and the energy consumption of
UE the IoT device.
Y1
According to [52] and [53], Lora is resistant to inter-
ference, due to its wide range of SFs, which can be used in
LoRa gw Y2 N2
N3IWF AMF urban, rural, and even industrial environments. Moreover, this
IoT dev technology offers four times the coverage range, compared to
Y2 other radio technologies due to its robust nature and ability
Y1
IoT dev
NWu
to work with low-intensity radio signals. Its ability to perform
N1 multiple transmissions on the same RF channel, with the use of
different SFs, reduces the likelihood of collisions, increasing the
effective data transmission rate, allowing discrimination between
time and frequency errors. Therefore, LoRa is considered one of
Fig. 20. Connections for the integration of untrusted non-3GPP access the most promising technologies for the physical layer of sensor
network with the 5G SBA core.
networks and, consequently, for IoT [54]. However, to have a
useful application of sensors and IoT, it is necessary to have a
3GGP does not define the Y1 connection, but this connection
network, such as LoRaWAN, described below.
establishes the communication of the device with its access
network, e.g., through a WiFi access point or LoRa gateway. 6.3 LoRaWAN network
It is assumed that there is some authorization process through
LoRaWAN network is the name given to IoT Low Power Wide
the Y1 connection, allowing the device to access the network
Area Network (LPWAN), which uses the LoRa technology as
and to obtain an IP address. However, several technologies, such
a physical medium and implement an architecture based on the
as LoRa, do not directly assign IP to the devices, and part of
LoRaWAN Media Access Control (MAC) protocol. In Fig. 21, we
the Y1 connection needs to be implemented at the technology
can see the MAC layer, its sublayers, and the LoRa physical layer,
gateway itself. The Y2 connection is also not defined by 3GPP,
enabling the long-range communication link. The MAC layer
but it must provide communication between the access network
protocol and the network architecture have a significant influence
and the N3IWF component of the SBA core. This connection
on determining the battery life a sensor, network capacity, QoS,
can be established in almost any way, including via the public
security, and the range of IoT applications served.
Internet, i.e., potentially with multiple equipment and intermediate
technologies.
The 3GPP standard specifies that between the device (or Application
gateway) and the N3IWF component, an encrypted IPSec tunnel,
called NWu, must be established for sending data and signaling
LoRa MAC
traffic. N3IWF selects AMF to serve the device, and the N2
interface between the two functions is established. After setting MAC layer
NWu and N2, the 3GPP standard specifies that it is possible to MAC options
create the N1 interface between devices (or gateway) and the AMF
component for transporting Non-Access Stratum (NAS) signaling. Class A Class B Class C
This approach differs from that used in the 4G EPC core, in
which NAS signaling applies only to 3GPP access networks. In the LoRa modulation
5G SBA core, devices connected via non-3GPP access networks
are managed similarly to devices connecting via 3GPP access Physical layer
ISG Regional Band (PHY)
networks.
EU US AS KR
-----
868 915 923 920
6.2 LoRa
LoRa technology, developed by Semtech Corporation [51], Fig. 21. LoRaWAN architecture layers.
allows communication over long distances using the Spread Spec-
trum concept of RF. This technology has the following con- The LoRaWAN protocol is based on open, low-cost standards.
figuration parameters that directly influence the communication This protocol was designed from the beginning to implement plat-
performance: forms for IoT. In this way, the scope of supported IoT applications
 20

is broad, making this technology attractive to the 5G system and MAC address, identifier (AppEUI) and application key (AppKey)
creating an IoT network with a large coverage area. Among the desired. This data is used in the application layer to validate and
possible uses, there are smart grids for electrical energy, sensor activate it in a given network application. After the acceptance
networks of several types, precision agriculture, and others. on the network, the device receives a message join accept,
Fig. 22 shows the LoRWAN protocol reference topology. A which has the device address (DevAddr), the network session
prominent feature of this technology is its low energy consump- key (NwkSKey), and the application session key (AppSKey). In
tion, due to the star topology that does not require routing between private networks, the following data are required for activation:
nodes. The central component is the LoRaWAN gateway that can device address (DevAddr), network session key (NwkSKey), and
be connected to a 5G system on untrusted non-3GPP access. These application session key (AppSKey), which are recorded on the
gateways, also known as concentrators, have models that can meet device, at the time of configuration. Therefore, the device is ready
specific demands, simpler and cheaper for closed environments, in for communication when it is connected to the network.
contrast to industrial models with protection against the weather In the LoRaWAN architecture, Network Servers are still re-
and external applications. In communication between gateway and sponsible for managing the information sent by gateways. As
devices, there are two previous types of messages exchange: Un- there is the possibility that two or more gateways receive the
confirmed Data Message, similar User Datagram Protocol (UDP) same packet from a given device, the network server eliminates
messages, and Confirmed Data Message, like to Transmission duplicate packets, manages the acknowledgment (ACK) return
Control Protocol (TCP). Another fundamental characteristic of times, and makes the adjustments for the adaptative data rate to
the LoRaWAN network is the existence of three devices classes manage the times between communications and energy consump-
that communicate with the gateways, which are described in the tion. Finally, LoRaWAN has one or more Application Servers that
following. receive packets from Network Servers via request or automatically
and, according to the request, perform one or more specific actions
IoT Concentrator/ Network Application providing the needed interface for various client applications.
sensors Gateway server server

Asset tracking
6.4 Demonstration of the integration of non-3GPP wire-
LoRA
less access network and the 5G core
Vending
machines
Goals
LoRA
The goal of the demonstration is to introduce a non-3GPP
Gas monitoring
wireless access network to the 5G core. Moreover, this experiment
LoRA
shows the great capillarity that the 5G system can have, including
network with unlicensed frequencies.
Garbage LoRa RF TCP/IP SSL TCP/IP SSL
monitoring
LoRaWAN LoRaWAN Secure Payload
Secure Payload AES Description
Application Data
NwkSkey
In this demonstration, we combine a RAN based on LTE
AppSkey
technology with a LoRaWAN wireless network implemented in
hardware. For RAN LTE, we use open-source software and an
SDR. We also use an open-source implementation of the SBA-
Fig. 22. LoRaWAN network topology based 5G core software. All components are implemented using
Docker containers that can be hosted on a cloud infrastructure.
Connected to access the network through LTE, there is a Lo-
• Class A: upstream transmissions (device-gateway) are RaWAN gateway implemented in generic hardware with sup-
based on the ALOHA protocol. The reception of the can port to multiples IoT sensors that synchronize its data with the
only be performed in two short reception windows that LoRaWAN server present at the other end of the infrastructure.
open after a transmission. This class provides the lowest Therefore, for demonstration purposes, the data collected by the
energy consumption, with the main application of the sensors are transmitted via the LoRa network to the gateway that
monitoring quantities. forwards the data via LTE backhaul, passing through the SBA core
• Class B: in addition to the class A mode, pre-programmed of the 5G network until reaching the LoRaWAN server. Fig. 23
transmission windows (gateway-device) and managed by a shows the experiment we designed to demonstrate the integration
timing signal (Beacon) are opened, which indicates when of a non-3GPP wireless access network with a 5G core network.
the receiver is ready to receive the signal. This class can
be convenient for remote control systems that are not time-
LoRaWAN LTE
sensitive. network network
LoRA UE
• Class C: the transmission window (gateway-device) is
LoRaWAN
open continuously, closing only at the time of the upstream Gateway with
IoT LTE backhaul LoRaWAN
transmission (device-gateway). sensors server
SDR
SBA
Regarding the activation of devices on the network, LoRaWAN
adopts the AES-128 encryption algorithm, using two different
forms of activation due to the Public or Private network. In public Fig. 23. Demonstration of the integration of a 5G core with a non-3GPP
networks, it uses Over The Air Activation (OTAA), which is based access network based on LoRa.
on sending a unique global identifier (DevEUI), analogous to the
 21

Additional information Security evolution

During the tutorial, we present a demonstration video of the Security has always been a concern in mobile wireless net-
experiment. Moreover, a manual is available with details on how works, especially from 3G with the broader use of infrastructure
the practices can be replicated. Finally, containers and any extra for data transport and its natural integration with the Internet.
code produced by the team needed to replicate the experiment is Therefore, the security issue in 5G networks has been addressed
also publicly available. from its design. It has been a recurring topic, mainly due to the
Repository for this tutorial: strategical importance of 5G in the economy and social issues.
https://github.com/LABORA-INF-UFG/NetSoft2020-Tutorial4. There is a list of security contributions planned for Release 16, of
which some relevant items are highlighted the following:
7 F INAL CONSIDERATIONS • Support for NPNs with new authentication schemes;
The academy has already started to investigate potential • Network slice specific authentication option, in addition to
challenges for the next generation of wireless mobile networks. primary authentication;
Naturally, there is still a lot of uncertainty, and several forecasts • Advanced security for Radio Resource Control and NAS
cannot be consolidated. However, some themes are beginning to be signaling;
addressed preliminarily in the 3GPP release scope. For example, • Support for integrity protection in the user plane, covering
Releases 16 and 17 improvements in URLLC and the use of SON the three 5G network scenarios, i.e., eMBB, URLLC e
to make the 5G networks more efficient are expected, indicated mMTC.
that forecasts such as MBRLLC and mURLLC, as well as, the
Other advances
broad adoption of Machine Learning and Artificial Intelligence,
can become a reality in the future. In the following, we briefly There are several other contributions in Release 16 and 17 that
discuss some topics addressed in the next two releases, scheduled introduce new solutions and offer opportunities for the research.
for June 2020 (Release 16) and September 2021 (Release 17). This opportunity occurs because specific solutions are not entirely
defined due to factors such as complexity of the problem addressed
or inability to identify in the standard the specifics that the answer
Data-driven network
will have to deal with that problem. In the following, we list some
The introduction of NWDAF in Release 15 is an essential topics of these two Releases (16 and 17) that, although they have
step for adopting ML and AL in a 5G network, but it is only the already been investigated, still present open questions, especially
beginning. The complete specification of the data collection and in a practical context and with the potential large-scale application.
analysis framework must be completed in Release 16. Form that
• Advances on Vehicle-to-everything (V2X) – platoon for-
point on, solutions based on ML and AI can use the information
mation, autonomous steering, and remote steering;
collected by NWDAF to perform tasks such as predicting and
• Access to unlicensed spectrum using 5G New Radio;
mobility optimization, detecting anomalies, QoS predicting, and
• Dynamic Spectrum Sharing (DSS);
data correction. In this context, some of the goals for Release 17
• IoT for Non-Terrestrial Networks (NTNs);
are:
• Support for Unmanned Aerial Systems (UAS).
• Predictable network performance assisted by NWDAF; Finally, we can see the increasingly strong integration between
• UE oriented data analysis; the areas of Telecommunications and Information Technology. On
• Expose of NWDAF data analysis for user applications; the one hand, the adoption of software and cloud-native model in
• NWDAF supporting the detection of anomalous events and 5G systems, i.e., RAN and core has grown. On the other hand, the
helping to analyze their causes. interest to create software solutions that meet the specific needs
In a long time, the objective is to use ML and AI techniques to of mobile wireless networks, such as edge computing and signal
automate network management with the minimum possible human processing systems, has also increased. This integration will still
intervention. It is fundamental to highlight that this management be explored mainly in the next years, to make development even
should involve different network types (i.e., heterogeneous net- more agile and interoperable. An example of this change in view
works) connecting to a common core based on the SBA model. is the distribution (from Release 15) of 3GPP specifications in
the format of readable data serialization (YAML - YAML Ain’t
Improvements for vertical domains Markup Language) [55], allowing the design and development
5G networks have been designed with particular attention to of standardized interfaces and with high flexibility. Moreover,
vertical domains (usually called only as verticals), i.e., specific the use of cloud-native technologies will allow extremely auto-
sectors or groups of companies in which similar products and mated management and control, with minimal (or even without)
services are developed, produced, and supplied. In this context, human intervention, which has been called Zero Touch Network
there are several directions and contributions in Release 16, for & Service Management (ZSM) [56]. These developments are
example: expected to change the business model of telecommunication
operators, allowing them to expand Business to Consumer (B2C)
• Support to Time Sensitive Communication (TSC); and Business to Business (B2B) relationships.
• Non-Public Networks (NPNs), i.e., private networks;
• Support to 5G LAN services; R EFERENCES
• Advanced location services. [1] M. Shafi, A. F. Molisch, P. J. Smith, T. Haustein, P. Zhu, P. De
In Release 17, improvements in support for NPNs and Prox- Silva, F. Tufvesson, A. Benjebbour, and G. Wunder, “5G: A Tutorial
Overview of Standards, Trials, Challenges, Deployment, and Practice,”
imity Services (ProSe) are planned, which are also useful in public IEEE Journal on Selected Areas in Communications, vol. 35, no. 6, pp.
security scenarios. 1201–1221, 2017.
 22

[2] A. Kaloxylos, “A Survey and an Analysis of Network Slicing in 5G [26] M. Giordani, M. Polese, M. Mezzavilla, S. Rangan, and M. Zorzi,
Networks,” IEEE Communications Standards Magazine, vol. 2, no. 1, “Toward 6G Networks: Use Cases and Technologies,” IEEE Commu-
pp. 60–65, 2018. nications Magazine, vol. 58, no. 3, pp. 55–61, 2020.
[3] F. Richter, “Global 5G Adoption to Take Off in 2021,” https://www. [27] Z. Chen, X. Ma, B. Zhang, Y. Zhang, Z. Niu, N. Kuang, W. Chen,
statista.com/chart/9604/5g-subscription-forecast/, 2019, [Último acesso: L. Li, and S. Li, “A survey on terahertz communications,” China
31-mar-2020]. Communications, vol. 16, no. 2, pp. 1–35, 2019.
[4] World Economic Forum, “The Impact of 5G: Creating New Value [28] O. Semiari, W. Saad, M. Bennis, and M. Debbah, “Integrated Millimeter
across Industries and Society,” https://www.weforum.org/whitepapers/ Wave and Sub-6 GHz Wireless Networks: A Roadmap for Joint Mobile
the-impact-of-5g-creating-new-value-across-industries-and-society, Jan Broadband and Ultra-Reliable Low-Latency Communications,” IEEE
2020, [Último acesso: 31-mar-2020]. Wireless Communications, vol. 26, no. 2, pp. 109–115, 2019.
[5] 3GPP, “4G; Release 8,” https://www.3gpp.org/specifications/releases/ [29] M. Peng, D. Liang, Y. Wei, J. Li, and H. Chen, “Self-configuration
72-release-8, 2009, [Último acesso: 26-fev-2020]. and self-optimization in LTE-advanced heterogeneous networks,” IEEE
[6] ——, “TR21.915: Technical Specification Group Services and System Communications Magazine, vol. 51, no. 5, pp. 36–45, 2013.
Aspects; Release 15 Description; Summary of Rel-15 Work Items [30] K. B. Letaief, W. Chen, Y. Shi, J. Zhang, and Y. A. Zhang, “The Roadmap
(Release 15),” https://portal.3gpp.org/desktopmodules/Specifications/ to 6G: AI Empowered Wireless Networks,” IEEE Communications
SpecificationDetails.aspx?specificationId=3389, 3rd Generation Partner- Magazine, vol. 57, no. 8, pp. 84–90, 2019.
ship Project (3GPP), Technical Report (TR) TR21.915, 2018-12, version [31] A. Adadi and M. Berrada, “Peeking Inside the Black-Box: A Survey
0.5.0. on Explainable Artificial Intelligence (XAI),” IEEE Access, vol. 6, pp.
[7] S. Abdelwahab, B. Hamdaoui, M. Guizani, and T. Znati, “Network 52 138–52 160, 2018.
function virtualization in 5G,” IEEE Communications Magazine, vol. 54, [32] A. Taleb Zadeh Kasgari, W. Saad, and M. Debbah, “Human-in-the-Loop
no. 4, pp. 84–91, 2016. Wireless Communications: Machine Learning and Brain-Aware Resource
[8] T. Chen, M. Matinmikko, X. Chen, X. Zhou, and P. Ahokangas, “Soft- Management,” IEEE Transactions on Communications, vol. 67, no. 11,
ware defined mobile networks: concept, survey, and research directions,” pp. 7727–7743, 2019.
IEEE Communications Magazine, vol. 53, no. 11, pp. 126–133, 2015. [33] M. A. Marotta et al., “Characterizing the Relation Between Processing
[9] X. Foukas, G. Patounas, A. Elmokashfi, and M. K. Marina, “Network Power and Distance Between BBU and RRH in a Cloud RAN,” IEEE
slicing in 5G: Survey and challenges,” IEEE Communications Magazine, Wireless Communications Letters, vol. 7, no. 3, pp. 472–475, 2018.
vol. 55, no. 5, pp. 94–100, 2017. [34] 3GPP, “TR38.801: 33rd Generation Partnership Project; Technical
[10] F. Mademann, “The 5G System Architecture,” Journal of ICT Standard- Specification Group Radio Access Network; Study on new ra-
ization, vol. 6, no. 3, pp. 77–86, 2018. dio access technology: Radio access architecture and interfaces
[11] M. R. Palattella, M. Dohler, A. Grieco, G. Rizzo, J. Torsner, T. Engel, and (Release 14),” https://portal.3gpp.org/desktopmodules/Specifications/
L. Ladid, “Internet of Things in the 5G Era: Enablers, Architecture, and SpecificationDetails.aspx?specificationId=3056, 3rd Generation Partner-
Business Models,” IEEE Journal on Selected Areas in Communications, ship Project (3GPP), Technical Report (TR) TR38.801, 2017-03, version
vol. 34, no. 3, pp. 510–527, 2016. 14.0.0.
[12] M. A. Lema, A. Laya, T. Mahmoodi, M. Cuevas, J. Sachs, J. Markendahl, [35] S. Gonzalez-Diaz et al., “Integrating fronthaul and backhaul networks:
and M. Dohler, “Business Case and Technology Analysis for 5G Low Transport challenges and feasibility results,” IEEE Transactions on
Latency Applications,” IEEE Access, vol. 5, pp. 5917–5935, 2017. Mobile Computing, pp. 1–18, 2019.
[36] F. Fonseca, S. Correa, and K. Cardoso, “Optimizing allocation and
[13] Recommendation ITU-R M.2083-0, “IMT Vision – Framework and
positioning in a disaggregated radio access network aware of paths
overall objectives of the future development of IMT for 2020
through the core infrastructure,” in Anais do XXXVII Simpósio Brasileiro
and beyond,” https://www.itu.int/dms pubrec/itu-r/rec/m/R-REC-M.
de Redes de Computadores e Sistemas Distribuı́dos, 2019, pp. 791–804.
2083-0-201509-I!!PDF-E.pdf, 2015.
[37] S. Karnouskos, A. W. Colombo, T. Bangemann, K. Manninen, R. Camp,
[14] U. N. Kar and D. K. Sanyal, “An overview of device-to-device commu-
M. Tilly, P. Stluka, F. Jammes, J. Delsing, and J. Eliasson, “A SOA-based
nication in cellular networks,” ICT Express, vol. 4, no. 4, pp. 203–208,
architecture for empowering future collaborative cloud-based industrial
2018.
automation,” in IECON 2012 - 38th Annual Conference on IEEE Indus-
[15] W. Stallings, Data and computer communications, 10th ed. Prentice trial Electronics Society, Oct 2012, pp. 5766–5772.
Hall, 2014.
[38] S. Imadali and A. Bousselmi, “Cloud Native 5G Virtual Network Func-
[16] J. H. Schiller, Mobile Communications, 2nd ed. Pearson Education, tions: Design Principles and Use Cases,” in 2018 IEEE 8th International
2003. Symposium on Cloud and Service Computing (SC2), Nov 2018, pp. 91–
[17] H. Holma and A. Toskala, LTE for UMTS – Evolution to LTE-Advanced, 96.
2nd ed. John Wiley & Sons, 2011. [39] P. Hedman, L. Frid, S. Rommer, M. Olsson, S. Sultana, and C. Mulligan,
[18] Recommendation ITU-R M.1645, “Framework and overall objec- 5G Core Networks: Powering Digitalisation. Elsevier Science &
tives of the future development of IMT-2000 and systems be- Technology, 2019.
yond IMT-2000,” https://www.itu.int/dms pubrec/itu-r/rec/m/R-REC-M. [40] 3GPP, “TS133.185: 3rd Generation Partnership Project Technical Spec-
1645-0-200306-I!!PDF-E.pdf, 2003. ification Group Services and Systems Aspects Release 14 Descrip-
[19] S. Rangan, T. S. Rappaport, and E. Erkip, “Millimeter-Wave Cellular tion,” https://www.etsi.org/deliver/etsi ts/133100 133199/133185/14.00.
Wireless Networks: Potentials and Challenges,” Proceedings of the IEEE, 00 60/ts 133185v140000p.pdf, 3rd Generation Partnership Project
vol. 102, no. 3, pp. 366–385, 2014. (3GPP), Technical Specification (TS) TS133.185, 2017-10, version
[20] X. Wang, L. Kong, F. Kong, F. Qiu, M. Xia, S. Arnon, and G. Chen, 14.1.0.
“Millimeter Wave Communication: A Comprehensive Survey,” IEEE [41] A. Toskala and M. Poikselkä, “5G Architecture,” 5G Technology: 3GPP
Communications Surveys Tutorials, vol. 20, no. 3, pp. 1616–1653, 2018. New Radio, pp. 67–86, 2020.
[21] M. Giordani, M. Polese, A. Roy, D. Castor, and M. Zorzi, “A Tutorial [42] ETSI, “TS 123 501: System architecture for the 5G System (5GS)(3GPP
on Beam Management for 3GPP NR at mmWave Frequencies,” IEEE TS 23.501 version 15.5.0 Release 15,” https://www.etsi.org/deliver/etsi
Communications Surveys Tutorials, vol. 21, no. 1, pp. 173–196, 2019. ts/123500 123599/123501/15.05.00 60/ts 123501v150500p.pdf, Euro-
[22] D. Zhang, A. Li, M. Shirvanimoghaddam, P. Cheng, Y. Li, and pean Telecommunications Standards Institute (ITSI), Technical Speci-
B. Vucetic, “Fast Beam Tracking for Millimeter-Wave Systems Under fication (TS) TS 123 501, 2019, version 15.5.0.
High Mobility,” in ICC 2019 - 2019 IEEE International Conference on [43] V. K. Choyi, A. Abdel-Hamid, Y. Shah, S. Ferdi, and A. Brusilovsky,
Communications (ICC), 2019, pp. 1–6. “Network slice selection, assignment and routing within 5G Networks,”
[23] Samsung Electronics Co, “5G Core Vision – Samsung 5G Core Vol.1,” in IEEE Conference on Standards for Communications and Networking
Samsung Electronics Co, Tech. Rep., 2019. (CSCN), 2016, pp. 1–7.
[24] F. Kaltenberger, G. d. Souza, R. Knopp, and H. Wang, “The OpenAirIn- [44] S. Parkvall, E. Dahlman, A. Furuskar, and M. Frenne, “NR: The new 5G
terface 5G New Radio Implementation: Current Status and Roadmap,” in radio access technology,” IEEE Communications Standards Magazine,
WSA 2019; 23rd International ITG Workshop on Smart Antennas, 2019, vol. 1, no. 4, pp. 24–30, 2017.
pp. 1–5. [45] A. Ghosh, A. Maeder, M. Baker, and D. Chandramouli, “5G Evolution:
[25] W. Saad, M. Bennis, and M. Chen, “A Vision of 6G Wireless A View on 5G Cellular Technology Beyond 3GPP Release 15,” IEEE
Systems: Applications, Trends, Technologies, and Open Research Access, vol. 7, pp. 127 639–127 651, 2019.
Problems,” CoRR, vol. abs/1902.10265, 2019. [Online]. Available: [46] 3GPP, “5G; Access to the 3GPP 5G Core Network (5GCN) via non-
http://arxiv.org/abs/1902.10265 3GPP access networks (Release 15),” https://www.etsi.org/deliver/etsi
 23

ts/124500 124599/124502/15.02.00 60/ts 124502v150200p.pdf, 2019,

[Último acesso: 03-fev-2020].
[47] WBA and NGMN Alliance, “RAN Convergence Pa-
per,” https://wballiance.com/wp-content/uploads/2019/01/
RAN-Convergence-Paper-2019.pdf, 2019, [Último acesso: 03-fev-
2020].
[48] J. Navarro-Ortiz, S. Sendra, P. Ameigeiras, and J. M. Lopez-Soler,
“Integration of LoRaWAN and 4G/5G for the Industrial Internet of
Things,” IEEE Communications Magazine, vol. 56, no. 2, pp. 60–67,
Feb 2018.
[49] R. Yasmin, J. Petajajarvi, K. Mikhaylov, and A. Pouttu, “On the inte-
gration of LoRaWAN with the 5G test network,” in 2017 IEEE 28th
Annual International Symposium on Personal, Indoor, and Mobile Radio
Communications (PIMRC), Oct 2017, pp. 1–6.
[50] M. Olsson, C. Mulligan, S. Sultana, S. Rommer, and L. Frid, EPC and 4G
packet networks: driving the mobile broadband revolution. Academic
Press, 2013.
[51] Semtech, “Semtech corporation,” https://www.semtech.com/, 2020,
[Último acesso: 22-abril-2020].
[52] J. P. Shanmuga Sundaram, W. Du, and Z. Zhao, “A survey on lora
networking: Research problems, current solutions, and open issues,”
IEEE Communications Surveys Tutorials, vol. 22, no. 1, pp. 371–388,
2020.
[53] JungWoon Lee, DongYeop Hwang, JiHong Park, and Ki-Hyung Kim,
“Risk analysis and countermeasure for bit-flipping attack in lorawan,” in
International Conference on Information Networking (ICOIN), 2017, pp.
549–551.
[54] J. Haxhibeqiri, E. De Poorter, I. Moerman, and J. Hoebeke, “A Survey of
LoRaWAN for IoT: From Technology to Application,” Sensors, vol. 18,
p. 3995, 11 2018.
[55] 3GPP, “Openapi,” feb 2020, [Último acesso: 26-abril-2020]. [Online].
Available: https://www.3gpp.org/FTP/Specs/archive/OpenAPI/
[56] ETSI, “Zero-touch network and Service Management (ZSM);
Terminology for concepts in ZSM,” European Telecommunica-
tions Standards Institute (ITSI), Technical Specification (TS),
2019, v1.1.1. [Online]. Available: https://www.etsi.org/technologies/
zero-touch-network-service-management

View publication stats