Building a 5G Core Network Testbed:

Open-Source Solutions, Lessons Learned,

and Research Directions
Linh-An Phan, Dirk Pesch, Utz Roedig, Cormac J. Sreenan
School of Computer Science and Information Technology
University College Cork, Cork, Ireland
[email protected],[email protected],[email protected],[email protected]

Abstract—Building a 5G Core Network (5GC) testbed is a RAN simulators and 5GC software without having expensive
useful exercise to gain a comprehensive understanding of 5G radio and end-device equipment. However, selecting and com-
technology and is an important part of 5GC research. How- bining these solutions into a comprehensive and functional
ever, this task requires significant engineering effort, making it
challenging, especially for novice researchers. In this paper, we software-based 5G testbed is a non-trivial task that demands
share insights from constructing an end-to-end software-based significant engineering effort, such as installing deployment
5G testbed, aiming to assist other researchers in overcoming environments and configuring networks [11].
technical challenges in this endeavour. We begin by providing Several industry-scale testbeds [12], [13] have been ini-
an overview of popular open-source software designed for 5G tiated recently in Europe and America, aiming to facilitate
testbeds, with a focus on the mobile core network. We then
present two distinct 5GC testbed designs: one employing two advanced research in mobile networks. Conducting research
desktop PCs and the other utilising a single high-performance on these testbeds offers significant benefits; however, this
server. Lessons learned during the testbed installation process are opportunity may not be accessible for many researchers. On
noted. Furthermore, we discuss several research opportunities the contrary, while building a lab-scale testbed may not be
that can be explored with the testbed, ranging from service a scientific contribution per se; this task is still worthwhile
development to automation in 5G deployments. We believe this
work will foster further exploration and experimentation within and offers two valuable benefits. First, it helps researchers to
the 5GC area. deeply understand the operation of the 5GC software system,
Index Terms—5G, Core Network, 5GC, testbed, open-source allowing them to discover a range of practical problems.
Secondly, it provides an empirical environment for testing
I. I NTRODUCTION research hypotheses as well as evaluating proposed solutions,
Research related to 5G technology can basically be clas- ensuring that the findings and the contribution are accurate,
sified into two broad topics: Radio Access Network (RAN) reliable, and reproducible. The evolution of the mobile core
and 5G Core Network (5GC), with the majority of studies network towards 5G has made it into a highly complex system.
concentrating on the first category. In both areas, testbeds Therefore, building and running a 5GC testbed is a practical
play an important role in facilitating empirical research stud- approach to start exploring research problems in 5G.
ies. Unlike the RAN, which requires obtaining a spectrum This paper shares our hands-on experience in the journey
license and specialised radio hardware, a 5GC testbed can of building a software 5GC testbed and is aimed especially
be installed on Commercial Off-The-Shelf (COTS) hardware. at researchers who may be new entrants to this field or have
This is because components of 5GC (i.e., network functions) limited computational resources. Our goal is to alleviate the
are fundamentally just applications that are designed to run intricacies of building a 5G testing environment and, in doing
on commodity servers [1]. There are two main approaches so, facilitate innovative research related to 5G, especially the
to deploying a 5GC testbed: 1) purchasing commercial 5GC 5GC. The contributions in this paper are three-fold:
• We first provide a comprehensive survey of open-source
software [2]–[5] or, 2) using open-source 5G solutions [6]–[9].
Although the former option offers better technical support, it is software that can be utilised to build a 5GC testbed. These
more expensive and offers less flexibility in terms of modifying software packages, namely OpenAirInterface 5G Core
the source code and sharing testbed implementations with the Network [6], Open5GS [7], free5GC [8], and Aether [9],
community [10]. Therefore, a majority of researchers tend to are still being actively developed and widely used in the
utilise 5G open-source projects for their research, allowing research community. We also cover several open-source
them to build an end-to-end 5G system using open-source RAN simulators and Open RAN solutions.
• We showcase two of our testbed designs, serving as
This publication has emanated from research conducted with the financial a reference for other researchers to set up their own
support of Science Foundation Ireland under Grant number 13/RC/2077 P2. testbeds. The first testbed uses two desktop PCs, while
For the purpose of Open Access, the author has applied a CC-BY public
copyright licence to any Author Accepted Manuscript version arising from the second one utilises a single, high-performance server.
this submission. Moreover, we also share setup guidelines and lessons

979-8-3503-3094-6/24/$31.00 ©2024 IEEE 644 ICOIN 2024

 that we learned during the creation of the testbeds,
providing insights from our experience, supplemented by
our GitHub repository.
• Lastly, we discuss several research directions that can
be undertaken with the testbed. These directions in-
clude developing 5G services and applications, designing
cloud-native 5G systems, developing high-performance
data plane functions, and achieving automation in 5G
deployments.
The 5GC is at the heart of a 5G system, facilitating 5G
standalone (SA) deployments. The 5G SA allows network
operators to unleash the full capabilities of 5G technology,
including ultra-low latency, network slicing, and network
Fig. 1. Overall Architecture of a 5G system.
function exposure. Therefore, promoting in-depth research and
development of 5GC is essential for its wide adoption. We
believe that insights from our experiences can assist other
to manage multiple UEs and PDU sessions concurrently. Re-
researchers in overcoming technical challenges inherent in
garding deployment options, OAI 5G CN has three deployment
5GC research.
options: minimal 5GC (use only AMF, SMF, and UPF), basic
The structure of this paper is as follows: Section II intro-
5GC (include authentications and data storage functions on
duces popular open-source solutions that can be used to build a
top of the minimal deployment), and slicing 5GC (include
5GC testbed. Then, we present the two designs of our testbeds
all functions of the 5GC). In terms of the UPF, OAI offers
and lessons learned during the process of building these in
a choice between the SPGW-U [14] (developed from 4G
Section III. Section IV outlines potential research avenues that
with additional features for 5G), VPP-UPF [15] (a high-
can be conducted using a 5GC testbed. Finally, we conclude
performance UPF relying on the Vector Packet Processor and
the paper and suggest future work in Section V.
Data Plane Development Kit technologies), and production-
II. O PEN -S OURCE S OLUTION grade UPF [6]. The OAI 5G CN can be deployed in various
environments, such as native installation on bare-metal servers
A. 5G Core Network or virtual machines, a Docker-based environment using Docker
A 5G system can be divided into two sub-systems known as Compose, and a cloud environment using Helm Chart on
RAN and 5GC, as shown in Fig. 1. The 5GC is designed based Kubernetes (OpenShift cluster). The development of OAI is
on a Service-Based Architecture (SBA), with a Control and contributed by members of the OpenAirInterface Software
User Plane Separation strategy to enable agility and scalability Alliance (OSA). OSA currently has 97 members worldwide,
in network deployment and operation. The SBA decouples including network operators, companies, universities and re-
the 5GC into independent services (i.e., network functions) search institutes. At the time of writing, the latest version of
that implement different functionalities [16]. The 5G Control OAI 5G CN was released in May 2023, accompanied by a
Plane consists of multiple network functions (NFs), such as the roadmap for future releases.
Access and Mobility Management Function (AMF), Session 2) Open5GS [7]: Open5GS is an open-source 5GC im-
Management Function (SMF), Session Management Function plementation written in the C language. It comes with a
(SMF), etc. In contrast, only one NF operates in the data plane, Web application for managing UE subscription information,
that is the User Plane Function (UPF). Since 5G core NFs are suitable for testing. The implementation of Open5GS is based
completely software-based applications, several open-source on the 3GPP specifications Release 17, with a rich set of
implementations of 5GC have been developed and are widely supported features such as IPv6 support, the capability to
used in the research community. The characteristics of these handle multiple PDU sessions, handover functions, and the
open-source 5GC solutions are summarised in Table I. integration of Voice over LTE and Voice over 5G New Radio.
1) OpenAirInterface [6]: OpenAirInterface 5G Core Net- Compared to other 5GC projects, Open5GS benefits from
work (OAI 5G CN) is an open-source implementation of the a larger contributor base, resulting in weekly commits to
5G Standalone (SA) Core Network based on 3GPP Release its repository and more frequent software releases. However,
16 specifications. The OAI 5G CN components have been Open5GS does not offer official releases of container images
extensively tested by professional testers using commercial for Open5GS’s components; thus, users have to build images
5G base stations (gNBs) with COTS User Equipment (UE), if they want to deploy them in container-based environments.
OAI O-RAN solution, and open-source RAN simulators. The 3) free5GC [8]: free5GC is another open-source 5GC
OAI 5G CN currently supports fundamental functions such as software developed in Go, adhering to the 3GPP Release
UE registration, de-registration, and session management. It 15 specifications. It operates under the Apache 2.0 license,
also supports features such as mobility, paging, and network allowing any individual or organization to leverage free5GC
slicing. A notable feature of the OAI 5G CN is its capability for commercial products. One unique network function of

645
 TABLE I
C OMPARISON OF OPEN - SOURCE 5GC SOFTWARE .

OpenAirInterface (OAI)
Open5GS [7] free5GC [8] Aether [9]
5G Core Network [6]
3GPP Specification Release-16 Release-17 Release-15 Release-15
Language C++ C Go Go
Database MySQL mongo mongo mongo
High-performance UPF Yes No No Yes
Latest Release 05/2023 10/2023 06/2023 09/2023
Container Image Release Yes No Yes Yes
Number of Contributors 10+ 82 32 20+
Commercial Commercial
Commercial
Open RAN (OAI O-RAN) Commercial Open RAN (SD-RAN)
Tested RAN Open RAN (OAI O-RAN)
Simulators (UERANSIM, Simulators (UERANSIM) Simulators (UERANSIM,
Simulators (UERANSIM)
gnbsim, My-5GRANTester) gNBSim)
Native Application Native Application Native Application
Deployment Environment Docker Compose Docker Compose Docker Compose Kubernetes
Kubernetes Kubernetes Kubernetes
Open-source License OAI Public License V1.1 GNU AGPL v3.0 Apache 2.0 Apache 2.0

free5GC compared to other open-source 5GC solutions is the B. Radio Access Network
non-3GPP Inter-Working Function (N3IWF). N3IWF connects The RAN consists of user equipment and base stations.
untrusted non-3GPP devices, such as Wi-Fi Access Points 5G base stations provide wireless communication to UEs
and LoRaWAN Gateways, to the 5GC [17]. Free5GC offers and connect them to the 5GC. There are two approaches
container image releases to facilitate its usage and provides to setting up a RAN in a 5G testbed: simulation-based and
a Docker Compose file for automated network function de- Open RAN-based. The first approach is easy and convenient
ployment. Additionally, the contributors of free5GC have as it uses simulation software (e.g., UERANSIM [18], my5G-
maintained a list of articles, studies, and tutorials for those RANTester [20], gnbsim [19], or gNB Simulator [21]) to
new to the platform. execute the RAN functions. These RAN simulators usually
focus on implementing signalling messages in the control
plane and the tunnelling protocol in the data plane, making no
4) Aether [9]: Aether—a project of the Open Networking difference between simulated and real devices from the 5GC
Foundation (ONF)—is an open-source 5G edge cloud platform point of view. Open-source RAN simulators are quite suitable
aimed at supporting deployments of private 5G networks in for studies focusing on the functionalities of 5GC. Moreover,
enterprise environments. Recently, Aether has been used as it is also easy to create a large number of UEs accessing a 5G
the 5G platform for large-scale experimental research projects network for large-scale scenarios. UERANSIM is currently
[9]. The Aether community consists of 44 members, including one of the most widely used open-source RAN simulation
network operators, companies, and universities. It consists software.
of several components, including SD-RAN (i.e., Open RAN The Open RAN approach, also known as Software Defined
solution), SD-Core (i.e., 5G core network), and an aggregated Radio, is to install open-source radio software on Universal
Aether Management Platform (AMP). The SD-Core project is Software Radio Peripherals. This approach is preferred to
developed based on Open Mobile Evolved Core and free5GC build large-scale, physical 5G testbeds. In the Open RAN
platforms to support 4G/LTE, 5G non-standalone, and 5G stan- architecture, a base station can be split into the Radio Unit
dalone deployments. Aether is highly scalable, spanning from (RU), the Distributed Unit (DU), and the Centralized Unit
small deployments for research and development purposes to (CU). Open RAN software, such as srsRAN [22], OpenAirIn-
large distributed private networks for Industry 4.0 use cases. terface [6], and SD-RAN [23], provide implementations of CU
Aether provides two automated deployment options: Aether and DU that can be deployed on general-purpose computing
OnRamp for deploying a 5GC on a Kubernetes cluster and servers. Open RAN is an important mobile network technology
Aether-in-a-Box (AiaB) for deploying a 5G system, including that enables multi-vendor interoperability, intelligence, pro-
a RAN simulator and a 5GC, on a single machine. AiaB grammability, and agility in the RAN.
provides an automated way to deploy Aether, making it easy III. T ESTBEDS & L ESSONS L EARNED
to run basic tests and validate the installation of components
in the 5GC. However, because Aether employs Kubernetes A. Testbed
and Helm as its deployment frameworks, modifying default We chose Aether for the 5GC software and UERANSIM
configurations requires a deep understanding of these complex for the RAN simulator to build our testbeds. Aether uses
frameworks. Kubernetes, the de-facto platform for running cloud-native

646
 Fig. 2. Configuration of Testbed 1 using two desktop PCs.
Fig. 3. Configuration of Testbed 2 using five virtual machines on a single
powerful server.
systems, as its default deployment environment. Choosing
Aether allows us to reuse testbed configurations for future
deployments on public cloud services, which can opt for
large-scale studies and testing scenarios. For the 5GC, we
utilised the AiaB package provided by Aether and applied
several modifications to make it operate in a multiple-node
environment. As stated, we use UERANSIM to simulate UEs
and gNBs. UERANSIM creates a distinctive virtual network
interface for each UE, allowing users to transmit data from
applications (e.g., ping, iPerf, browser, media streaming) in the
5G data plane under diverse evaluation scenarios. Our setup
includes two distinct testbeds, namely Testbed 1 and Testbed
2, which are described below:
1) Testbed 1: Testbed 1 consists of two desktop PCs
running Ubuntu 22.04, as illustrated in Fig. 2. The first PC,
powered by an Intel Core i3 processor with 8GB RAM, runs
UERANSIM within a Docker Compose environment. The
second PC is equipped with an Intel Xeon processor and 12
GB of RAM. This PC is used to run the 5GC implementation
in the Kubernetes environment. Each PC is furnished with two
network interfaces, creating two separate connections between
the PCs. One network is configured for internet access and
control plane communication between RAN and 5GC, while
the other handles data plane communication. As NFs in the
5GC are deployed as a pod in Kubernetes, the communication
among NFs is handled by Kubernetes networking. Testbed 1
was initially designed for educational purposes and is used to
assist students in understanding 5G systems and 5GC. Despite
its setup using modest PCs, the computing capacity of Testbed
1 can be increased by adding more RAN nodes or deploying
the 5G control plane and user plane functions in different
machines. Fig. 4. (a) Network functions of 5GC are deployed in a Kubernetes
environment, (b) ping test from UE to a server on the Internet, and (c)
2) Testbed 2: Testbed 2 is an enhanced version of Testbed throughput test between a UE and a local virtual machine using iPerf3 in
1, comprising of five virtual machines (VMs) running on a Testbed 2.
single powerful server, as illustrated in Fig. 3. This testbed is
used in our ongoing research to demonstrate the feasibility
and advantages of the 5G LAN-type service in industrial to Testbed 1, RAN, 5GC, and Multi-access Edge Computing
applications [24]. The server is equipped with two Intel Xeon (MEC) servers are deployed in distinct networks (i.e., subnets).
processors (40 Core CPU) and 128 GB of RAM. In contrast Therefore, two virtual routers deployed in VM 2 and VM 4

647
are used to connect these networks together. Fig. 4(a) shows IV. R ESEARCH D IRECTIONS
that network functions of Aether are deployed as pods in a A. Cloud-Native 5G System
Kubernetes cluster. Meanwhile, Fig. 4(b) and 4(c) show a ping
Using a cloud-native architecture (i.e., SBA) is the ma-
test from a UE to a server on the Internet and a throughput test
jor difference between the 5G and 4G core networks [26].
between a UE and a local VM using the iPerf3 tool in Testbed
Although this software architecture offers multiple benefits,
2. It should be noted that VMs in Testbed 2 can be deployed
designing a 5GC cloud-native system is challenging. This task
on different physical servers. Moreover, adding more VMs to
involves refactoring network functions to independent services
each part of the 5G system will not alter the overall network
using container-based technology [27]. In a cloud-native 5G
configuration of the testbed. In other words, the design of
system, the failure of an individual service should not disrupt
Testbed 2 can support both vertical and horizontal scaling.
the operation of other services and cause a system-wide failure.
Detailed installation guidelines for both testbeds are available
Moreover, adhering to the 3GPP standards for cloud-native
in our GitHub repository [25].
design ensures cross-platform compatibility, enabling multi-
B. Lessons Learned vendor developments of NFs. In other words, agility, scala-
bility, and dependability are key characteristics of a cloud-
1) Deployment Platform: As 5G is moving to cloud-native
native system. However, existing open-source 5GC solutions
systems, installing a 5G testbed in a container-based envi-
have not entirely met these cloud-native prerequisites, making
ronment is needed. Docker Compose and Kubernetes are two
the design and development of cloud-native 5G systems an
popular container orchestration frameworks that can be used
ongoing research question. Employing a 5G testbed enables
for this purpose. Docker Compose is easy to use as it allows
researchers to implement and validate their solutions regarding
deploying the entire system with just a single command using
this research question.
a single configuration file. OpenAirInterface and free5GC
are 5GC software implementations that provide a Docker B. 5G Services & Applications
Compose file to install their components quickly. However, The advance of 5G technologies has led to the emergence of
Docker Compose is designed to run multiple containers on new services and applications. Many of them are developed on
a single machine (i.e., single-host deployment), which limits top of the 5GC, such as network slicing [28], 5G LAN-type
the scalability of the system. On the contrary, Kubernetes service [29], edge computing, and traffic steering [30]. The
can deploy and manage containers across multiple nodes (i.e., functionalities of fundamental services, such as device authen-
cluster deployment). Moreover, Kubernetes supports advanced tication and management, mobility management, and session
capabilities such as automatic scaling, load balancing, self- management, are also implemented in the 5GC. A 5G testbed
healing, and resource management. As a result, Kubernetes is allows researchers to verify their enhancements for existing
complex and time-consuming to use effectively. It is worth services or demonstrate the benefits of novel applications in
noting that Docker Compose and Kubernetes can be used a 5G environment. Another important research topic in 5G is
together in certain environments. For example, in our testbed, security, which can be conducted based on the actual operation
we use Docker Compose to simplify the deployment of the of 5G components in a confined and controlled testbed [31].
RAN while utilising Kubernetes to enable scalability in the Lastly, the 5G testbed allows researchers to generate a large
deployment of the 5GC. amount of operating data from devices in 5G networks, which
2) Network Configuration: Understanding the network of a can be used as realistic datasets for machine learning studies.
5G system and how it aligns with the network model of con-
tainer platforms is crucial to building a 5G testbed. Basically, C. High Performance Data Plane
communications in a 5G testbed can be divided into control The network performance of 5G requires a UPF to have
plane and data plane paths. It is recommended to use different high throughput and low latency in packet processing. Data
network interfaces (or even different physical networks) for plane acceleration solutions have been investigated in the last
each plane to avoid conflicting configurations. Moreover, since decade. Several technologies, such as Data Plane Development
we do not deploy the RAN in the Kubernetes cluster, the Kit (DPDK), Vector Packet Processing (VPP) [15], Express
communication between RAN and 5GC (or between Internet Data Path (XDP) [32], and P4 [33], have been utilised to
and UPF) is between the outside of the Kubernetes cluster develop high-performance UPFs. However, these technolo-
and components inside the cluster. This differs from the gies require specific hardware and configurations, which are
communication among network functions in the 5GC, which difficult to deploy in a container-based environment [34].
happens inside the Kubernetes cluster. Some NFs, such as Therefore, supporting high-performance, cloud-native UPF in
AMF, SMF, and UPF, require multiple network interfaces for existing 5GC solutions is crucial for the practical application
their operations. Moreover, most open-source 5GC implemen- of open-source 5GC software.
tations currently provide a single-machine deployment based
on a localhost network, which often does not require routing D. Automation
configuration. As a result, these sample deployments may not The complexity of 5G systems requires automated de-
operate accurately in multiple-node or cloud environments, ployment solutions to minimise human errors and reduce
which require additional network configurations. installation time. Additionally, the geographical distribution

648
and diverse network demands highlight the need for auto- [12] S. Fdida et al., “SLICES, a scientific instrument for the networking com-
mated deployment of the 5GC [35]. For example, UPFs and munity,” Computer Communications, vol. 193, pp. 189–203, September
2022.
application servers can be deployed at the network edge [13] T. Chen et al., “A Software-Defined Programmable Testbed for Beyond
in ultra-low latency communication systems, while control 5G Optical-Wireless Experimentation at City-Scale,” IEEE Network, vol.
plane functions reside on centralised cloud servers. Providing 36, no. 2, pp. 90–99, March/April 2022.
[14] OpenAirInterface Software Alliance, “SPGW-U: An
automated deployment solutions further enables researchers implementation of the Evolved Packet Core network,”
to quickly set up new testbeds and replicate previous research https://github.com/OPENAIRINTERFACE/openair-spgwu-tiny
results, facilitating more studies in the 5G domain. (accessed Oct. 23, 2023).
[15] travelping, “User Plane Gateway (UPG) based on VPP,”
https://github.com/travelping/upg-vpp (accessed Oct. 23, 2023).
V. C ONCLUSION & F UTURE W ORK [16] 3GPP TS 23.501, “System architecture for the 5G System (5GS),”
v18.2.2, Jul. 2023.
The 5G Core Network (5GC) is integral to 5G Standalone [17] M. T. Lemes, A. M. Alberti, C. B. Both, A. C. De Oliveira Júnior
deployments, allowing network operators to harness the full and K. V. Cardoso, “A Tutorial on Trusted and Untrusted Non-3GPP
potential of 5G technology. In this paper, we provided a Accesses in 5G Systems—First Steps Toward a Unified Communications
Infrastructure,” IEEE Access, vol. 10, pp. 116662–116685, Nov. 2022.
review of available open-source packages and shared our [18] UERANSIM, “UERANSIM GitHub Repository,”
experience building software-based 5GC testbeds that can https://github.com/aligungr/UERANSIM (accessed Oct. 23, 2023).
be deployed on commodity servers. Notably, these testbeds [19] Rohan Kharade, “gnbsim: a 5G SA gNB/UE Simulator for Testing
5G System,” https://gitlab.eurecom.fr/kharade/gnbsim (accessed Oct. 23,
do not require expensive hardware, making them ideal for 2023).
researching, developing, and testing with 5GC functionalities. [20] Lucas B.D. Silveira et al., “Tutorial on communication between access
It is important to emphasise that the scale of the testbed can be networks and the 5G core,” Computer Networks, vol. 216, pp. 109301,
October 2022.
tailored to adapt to various research scenarios. In future work, [21] Open Networking Foundation, “gNB Simulator GitHub Repository,”
we plan to deploy a 5GC testbed on cloud services, actualising https://github.com/omec-project/gnbsim (accessed Oct. 23, 2023).
a truly cloud-native 5GC deployment. Such an integration [22] Software Radio Systems, “srsRAN Project: Open Source RAN,”
https://www.srsran.com (accessed Oct. 23, 2023).
would enhance the scalability and flexibility of the testbed [23] Open Networking Foundation, “SD-RAN Project,”
and broaden its accessibility to the research community. We https://opennetworking.org/open-ran/ (accessed Oct. 23, 2023).
believe that our insights and efforts will facilitate further 5GC [24] L.-A. Phan, D. Pesch, U. Roedig and C. J. Sreenan, “Demonstration of
LAN-type Communication for an Industrial 5G Network,” In Proc. of
research and innovations. IEEE Conference on Computer Communications (INFOCOM), Hobo-
ken, NJ, USA, 2023.
ACKNOWLEDGMENT [25] Linh-An Phan, “5G LAN Testbed GitHub Repository,”
https://github.com/linhanphan/5glan (accessed Oct. 23, 2023).
Testbed 1 was mainly installed by Alban Guilhot (University [26] L. Peterson, O. Sunay, and B. Davie, “Private 5G: A Systems Approach,”
of Clermont Auvergne, France) under the author’s supervision https://5g.systemsapproach.org (accessed Oct. 23, 2023).
[27] A. Khichane, I. Fajjari, N. Aitsaadi and M. Gueroui, “Cloud Native 5G:
during his internship at University College Cork. an Efficient Orchestration of Cloud Native 5G System,” NOMS 2022-
2022 IEEE/IFIP Network Operations and Management Symposium,
R EFERENCES Budapest, Hungary, pp. 1–9, June 2022.
[28] S. Zhang, “An Overview of Network Slicing for 5G,” IEEE Wireless
[1] W. d. S. Coelho, A. Benhamiche, N. Perrot and S. Secci, “Network Communications, vol. 26, no. 3, pp. 111–117, June 2019.
Function Mapping: From 3G Entities to 5G Service-Based Functions [29] X. Zhu and Y. Chi, “5G Release 16 Key Technologies for Vertical
Decomposition,” IEEE Communications Standards Magazine, vol. 4, no. Industries Application,” 2021 International Wireless Communications
3, pp. 46-52, September 2020. and Mobile Computing (IWCMC), pp. 1269–1274, Aug. 2021.
[2] Amantya, “5G Lab: End-to-end Solution for R&D, Testing, and Use [30] A. K. Salkintzis, M. Kühlewind, S. Rommer and R. Liebhart, “Multi-
Case Development,” https://www.amantyatech.com/5g/5g-use-case-lab- path QUIC for Access Traffic Steering Switching and Splitting in 5G
setup (accessed Oct. 23, 2023). Advanced,” IEEE Communications Standards Magazine, vol. 7, no. 1,
[3] Athonet, “Athonet 5G Core,” https://athonet.com/products/athonet-5g- pp. 48–56, March 2023.
core (accessed Oct. 23, 2023). [31] M. Repetto, “Service Templates to Emulate Network Attacks in Cloud-
[4] Affirmed, “Affirmed UnityCloud—5G Core,” Native 5G Infrastructures,” 2023 IEEE 9th International Conference on
https://www.affirmednetworks.com (accessed Oct. 23, 2023). Network Softwarization (NetSoft), Madrid, Spain, 2023.
[5] L. R. Salvador and Z. Rajnai, “5G Vendors, cybersecurity and NESAS,” [32] T. A. N. do Amaral, R. V. Rosa, D. F. C. Moura and C. E. Rothenberg,
2023 IEEE 17th International Symposium on Applied Computational “An In-Kernel Solution Based on XDP for 5G UPF: Design, Prototype
Intelligence and Informatics (SACI), Timisoara, Romania, 2023. and Performance Evaluation,” 2021 17th International Conference on
[6] OpenAirInterface Software Alliance, “OpenAirInterface,” Network and Service Management (CNSM), Izmir, Turkey, 2021.
https://openairinterface.org (accessed Oct. 23, 2023). [33] R. MacDavid et al., “A P4-based 5G User Plane Function,” In Proc.
[7] Sukchan Lee, “Open5GS,” https://open5gs.org (accessed Oct. 23, 2023). of the ACM SIGCOMM Symposium on SDN Research (SOSR), USA,
[8] free5GC, “free5GC–Open source 5G core network base on 3GPP R15,” 2021.
https://free5gc.org (accessed Oct. 23, 2023). [34] S. Sekigawa, C. Sasaki and A. Tagami, “Toward a Cloud-Native Telecom
[9] Open Networking Foundation, “Aether: Private 5G Platform,” Infrastructure: Analysis and Evaluations of Kubernetes Networking,”
https://opennetworking.org/aether (accessed Oct. 23, 2023). 2022 IEEE Globecom Workshops (GC Wkshps), Rio de Janeiro, Brazil,
[10] D. Pineda, R. Harrilal-Parchment, K. Akkaya, A. Ibrahim and A. pp. 838–843, Dec. 2022
Perez-Pons, “Design and Analysis of an Open-Source SDN-based 5G [35] A. Santos, J. Bernardino and N. Correia, “Automated Application De-
Standalone Testbed,” IEEE INFOCOM 2023 - IEEE Conference on ployment on Multi-Access Edge Computing: A Survey,” IEEE Access,
Computer Communications Workshops (INFOCOM WKSHPS), Hobo- vol. 11, pp. 89393–89408, 2023.
ken, NJ, USA, 2023.
[11] D. Scotece, A. Noor, L. Foschini and A. Corradi, “5G-Kube: Complex
Telco Core Infrastructure Deployment Made Low-Cost,” IEEE Commu-
nications Magazine, vol. 61, no. 7, pp. 26–30, July 2023.

649