Network Slicing on 3GPP 5G System

Architecture; An End to End Approach

MohamadAli BayatMokhtari Nasser Yazdani
School of Electrical and Computer Engineering, School of Electrical and Computer Engineering,
College of Engineering, University of Tehran, College of Engineering, University of Tehran,
Tehran, Iran Tehran, Iran
[email protected] [email protected]

Received: 23 July 2022 – Revised: 25 November 2022 - Accepted: 12 March 2023

Abstract—New emerging industries, such as vertical markets, need diverse networking requirements that the next
generation mobile networks have to support effectively. Network slicing is the basic solution to meet the diverse
requirements of various services over a common network infrastructure. Different network slicing architectures have
been proposed; however, to the best of our knowledge, there is no unified architecture to cover all aspects of technology.
In this paper, we propose a complete network slicing architecture based on 3GPP 5G system that addresses a unified
end-to-end approach. We show how this architecture can create and operate various slices in the core and radio access
sections using SDN controllers, virtualization, NFV MANO, and 3GPP management functions. We do compare our
proposed one with some famous network-slicing architectures. The Comparison Results show that our proposed
architecture is complete and covers all the aspects of network slicing. It uses NFV management and orchestration
capabilities and SDN controllers while being compatible with 3GPP Service-Based architecture. It also provides life
cycle management of network slices in both creation and operation phases in both core and radio access domains of the
5G network. In addition, we have included a functional RAN layer split to lay the corresponding layers in centralized
or distributed units according to the requirements of each eMBB, mMTC, or URLLC slice.

Keywords: Network Slicing, 5G, SDN, NFV, Management and Orchestration, System Architecture
Article type: Research Article
© The Author(s).
Publisher: ICT Research Institute
architectural approach, are unable to address the
I. INTRODUCTION diverging performance requirements that verticals need
New business and vertical markets demand very in terms of bandwidth, latency, scalability and
diverse communication requirements. Communication reliability [3]. Therefore, 5G mobile network must have
services usually are classified into three main enough flexibility and scalability to create and operate
categories, enhanced mobile broadband (eMBB), multiple logical networks each tailored for a specific
massive machine-type communications (mMTC) and use case running simultaneously on a common physical
ultra-reliable and low-latency communications infrastructure in a flexible, agile, and cost-efficient
(URLLC) [1]. Legacy mobile networks are manner.
characterized by monolithic network elements that have Network slicing is key for this end. It is the
tightly coupled hardware, software and functionality capability that enables us to deploy multiple end-to-end
[2]. This networks, with their “one-size-fits-all”


Corresponding Author
16 Volume 15- Number 4 – 2023 (15 -24)

logical networks (slices) over a common physical management and orchestration entity. The concept of
infrastructure each fitted for a specific use-case [4]. network slicing is described by NGMN in [4]. The
Network slice is a composition of adequately network slicing concept consists of 3 layers: 1) Service
configured network functions, network applications, Instance Layer, 2) Network Slice Instance Layer, and 3)
and the underlying cloud infrastructure (physical, Resource layer. 5GPPP architecture working group in
virtual or even emulated resources, RAN resources [14], presented an overall 5G architecture. The
etc.), that are bundled together to meet the requirements recursive model is structured in three levels including
of a specific use case, e.g., bandwidth, latency, service level, Network level, and resources level. A
processing, and resiliency, coupled with a business secure network and service management entity
purpose [5]. Network softwarization can provide the manages the total system to provide end to end services.
programmability, flexibility, and modularity that is In [2], functional perspective of the overall 5G
required to create network slices. Software-Defined NORMA network slicing architecture is presented. It
Networking (SDN) and Network Function shows the separation into four layers (service layer,
Virtualization (NFV) technologies are two key enablers management and orchestration layer, control layer, and
to achieve network slicing in 5G networks [6], [7]. data layer) as well as the differentiation into intra-slice
and interslice functions. The Management &
Although a lot of works about network slicing and Orchestration (MANO) Layer realizes 5G NORMA’s
specially using SDN and NFV have been done that are Software-defined Mobile network Orchestration
reviewed in [8]–[12], but only a few of those have (SDMO) concept by extending the ETSI NFV
focused on total architectural framework that address management and orchestration (NFV MANO)
all aspects of network slicing. In this work, we propose architecture towards multi-tenant and multi-service
a complete network slicing architecture based on 3GPP networks. The Control Layer accommodates the two
5G system. We have included 3GPP network slicing main controllers: (1) the Software-Defined Mobile
concepts as well as ETSI NFV management and Network Coordinator (SDM-X) for the control of
orchestration (MANO) block in combination with SDN common (shared) NFs and (2) Software-Defined
controllers. Our proposed architecture provides Mobile Network Controller (SDM-C) for dedicated
different functional split options for radio access stack NFs.
protocol. This capability enables us to deploy RAN sub-
layers in central unit (CU or BBU) or in distributed unit 3GPP System Architecture for 5G is defined as
(DU or RRH) for any slice type based on its slice- service-based and the interaction between network
specific requirements. We show how this architecture functions [15]. It is represented in two ways: service-
can create and operate various MBB, mMTC, and based representation and reference point representation.
uRLLC slices in the core and radio access sections As our proposed architecture is based on 3GPP SA, it is
using SDN controllers, virtualization, NFV MANO, detailed in the following sections.
and 3GPP management functions.
III. TYPICAL MOBILE NETWORK
The rest of the paper is organized as follows. In the
next section, we briefly discuss related works. Then a A mobile network consists of two parts: the core
typical mobile network is presented, and 3GPP 5G network (CN) and radio access network (RAN).
system architecture is discussed. After that, we explain Network services are implemented in the CN, and the
the network slicing concept and network slice lifecycle RAN allows the user to connect to the network. The CN
management. Following that, we present a 5G network of 5G (5GC) functionalities are implemented as a set of
slicing architecture based on 3GPP SA and an NFV virtual network functions on core cloud servers. The 5G
Based Network Slice Management and Orchestration RAN (also NG-RAN) is connected to the CN via
architecture. Then, our proposed SDN/NFV based Backhaul links. In the new radio access networks, some
architecture for 5G network slicing is discussed. functional parts are separated from the base station (BS)
Finally, comparison results are presented. and they are implemented in a centralized part as virtual
functions on edge cloud servers. In this case, the
II. RELATED WORKS centralized section (CU) is called baseband unit (BBU)
and the distributed section (DU), which is usually the
Network slicing is defined by multiple SDOs and antenna and the radio side, is called the remote radio
Fora such as ONF, ETSI, NGMN and 3GPP. However, head (RRH). RRHs are connected to BBU via front haul
the meaning and understanding of the network slicing links. Fig. 1 shows a typical mobile network.
concept are different in each technology and there is no
a common definition and understanding. In [7], ETSI
NFV provides an insight into the different views about
network slicing and describes the possible relationship
with the NFV constructs. ONF in [6] presents a core
concepts of the SDN Architecture that can be applied to
5G network slicing. The hart of the SDN architecture is
the SDN controller. It consists of client and server
contexts in addition to virtualization and orchestration
functions. In [13], NGMN envisions a 5G architecture
that leverages the structural separation of hardware and
software, as well as the programmability offered by
SDN and NFV. The architecture comprises three layers
(business application layer, business enablement layer,
and infrastructure resource layer) and an E2E Fig. 1. Typical mobile network.
Volume 15- Number 4 – 2023 (15 -24) 17

IV. 3GPP 5G SYSTEM ARCHITECTURE • User Plane Function (UPF)

The 3GPP 5G System architecture is defined as • Application Function (AF)
Service-Based Architecture (SBA) [15]. SBA is • User Equipment (UE)
specified by Network Function Services and Service- • (Radio) Access Network ((R)AN
Based Interfaces. This architecture support data • Network Data Analytics Function (NWDAF)
connectivity and services enabling deployments to use
techniques such as Network Function Virtualization
A. 3GPP 5G NR
and Software Defined Networking. Network Functions
provide Network Services and expose them to other Overall description of 5G radio access network
Network Functions through Service-Based interfaces. (NG-RAN or 5G-NR) is provided by 3GPP in TS
The interaction between Network Functions is 38.300 [16]. Fig. 3 shows the NG RAN architecture:
represented in two ways (Fig. 2): B. NG Interface
• A service-based representation, where network An NG-RAN node is either a gNB, providing NR
functions within the Control Plane enables other user plane and control plane protocol terminations
authorized network functions to access their towards the UE; or an ng-eNB, providing E-UTRA user
services. plane and control plane protocol terminations towards
• A reference point representation, shows the the UE. The gNBs and ng-eNBs are interconnected with
interaction exist between the NF services in the each other by means of the Xn interface. The gNBs and
network functions described by point-to-point ng-eNBs are also connected by means of the NG
reference point between any two network interfaces to the 5GC, more specifically to the AMF by
functions. means of the NG-C interface and to the UPF by means
The most important network functions of 5G of the NG-U interface. Fig. 4 shows NG user plane and
System architecture are: control plane protocol stack [16].
• Authentication Server Function (AUSF) C. Radio Protocol Architecture
• Access and Mobility Management Function Fig. 5 shows the user plane and control plane 5G
(AMF) NR protocol stack which is used to connect UE to gNB.
• Data Network (DN) PHY, MAC, RLC, and PDCP sublayers are in both data
• Network Exposure Function (NEF) and control layers. SDAP is added to user plane to
• Network Repository Function (NRF) control QOS. Finally, RRC and NAS are exclusive
• Network Slice Selection Function (NSSF) control plane sublayers. The Functionality and services
• Policy Control Function (PCF) of this sublayers is described in [16].
• Session Management Function (SMF)
• Unified Data Management (UDM)
• Unified Data Repository (UDR)

Fig. 3. NG-RAN overall architecture [16]

Fig. 4. NG user plane and control plane protocol stack [16]

Fig. 2. 3GPP 5G SA. (a) Reference point representation. (b)
Service-based representation [15]
18 Volume 15- Number 4 – 2023 (15 -24)

Fig. 5. 5G NR user plane and control plane protocol stack [16]

V. 5G MOBILE NETWORK SLICING Fig. 6. 5G network slicing architecture based on 3GPP SA

Network slicing is creating multiple end-to-end
logical networks (slices) over the top of a common A. Network slice lifecycle management
physical network, each tailored for a specific use case. Management aspects of network slicing can be
Network slices are defined as end-to-end (E2E) logical described by four phases shown in Fig. 7 [18]. Each
networks running on a common underlying (physical or phase defines high level tasks and should include
virtual) network, mutually isolated, with independent appropriate verification of the output of each task.
control and management, which can be created on
demand [3]. Each network slice consists a set of • Preparation: In the preparation phase the Network
resources (computing, storage, and networking) that are Slice instance does not exist. The preparation
properly combined to satisfy service requirements. phase includes network slice design, network slice
capacity planning, on-boarding and evaluation of
Adopted from [17] we introduced a network slicing the network functions, preparing the network
architecture. Fig. 6 shows this architecture that is based environment and other necessary preparations
on 3GPP SA. Network slicing is done in both CN and required to be done before the creation of a
RAN. A complete E2E network slice instance (NSI) is Network Slice instance.
a combination of some network subnet slice instances • Commissioning: Network Slice instance
(NSSIs). Slicing in the core network segment is provisioning in the commissioning phase includes
straightforward due to its network function-based creation of the Network Slice instance. During
architecture. It is needed to create and chain a set of Network Slice instance creation all needed
network function instances for each slice instance, resources are allocated and configured to satisfy
depending on the operational and performance the network slice requirements. The creation of a
requirements. For overall network management, some Network Slice instance can include creation
functions on the control plane such as AMF and NSSF and/or modification of the Network Slice instance
can be shared within all slides. Other functions that are constituents.
slice-specific are deployed exclusively for each slice. • Operation: The Operation phase includes the
Each network slice may connect to a different external activation, supervision, performance reporting
data network (DN) or all connect to single one. (e.g. for KPI monitoring), resource capacity
However, this architecture covers only operation phase planning, modification, and de-activation of a
that network slice instances and their corresponding Network Slice instance. Activation makes the
network slice subnet instances (Here 3 typical eMBB, Network Slice instance ready to support
uRLLC, and mMTC Slices) have been already created communication services. Resource capacity
by the slice provider and are operating by service planning includes any actions that calculates
providers (slice operators). resource usage based on a Network Slice instance
The AMF instance that is serving the UE may be provisioning, and performance monitoring and
shared between slices. Other network functions, such as generates modification polices as a result of the
the SMF or the UPF, may be specific to each Network calculation. Network Slice instance modification
Slice. The Network Slice instance selection for a UE is could be including e.g. capacity or topology
normally triggered as part of the registration procedure changes. The modification can include creation or
by the first AMF that receives the registration request modification of Network Slice instance
from the UE. AMF may query the UDM to retrieve UE constituents. Network Slice instance modification
subscription information. The AMF retrieves the slices can be triggered by receiving new network slice
that are allowed by the user subscription and interacts requirements or as the result of
with the NSSF to select the appropriate Network Slice supervision/reporting. The deactivation includes
instance e.g., based on Allowed S-NSSAIs, PLMN ID, actions that make the Network Slice instance
etc. The establishment of a PDU session within the inactive and stops the communication services.
selected instances NSSAI is triggered when the AMF Network slice provisioning actions in the
receives a Session Management message from UE. The operation phase involves activation, modification
AMF discovers candidate Session Management and de-activation of a Network Slice instance.
Functions (SMF) using multiple parameters including • Decommissioning: Network Slice instance
the S-NSSAI provided in the UE request and selects the provisioning in the decommissioning phase
appropriate SMF. The selection of the UPF is includes decommissioning of non-shared
performed by the SMF and uses the S-NSSAI. The S- constituents if required and removing the Network
NSSAI associated with a PDU Session is provided to Slice instance specific configuration from the
the AN, and to the PCF entities, to apply slice specific shared constituents. After the decommissioning
policies. phase, the Network Slice instance is terminated
and does not exist anymore.
Volume 15- Number 4 – 2023 (15 -24) 19

slice subnets consisting of physical and or virtual

network functions in both CN and RAN sections. It also
includes connectivity between NFs.
3GPP TR 28.801 [21] describes an information
model where a network slice contains one or more
Fig. 7. Management aspects of network slicing [18] network slice subnets, each of which in turn contains
one or more network functions and can also contain
B. 3GPP Management and Orchestration for other network slice subnets. These network functions
Network Slicing can be managed as VNFs and/or PNFs. To address
Network function virtualization (NFV) is based on management and orchestration of communication
decoupling of network functionality from hardware services, 3GPP introduces Management Function (MF)
infrastructure. Instead of network equipment (NE), its and Management Service (MS) concepts in [22]. A
Functional task is considered as network function (NF). management service offers management capabilities. In
Multiple instances of network functions can be created the service-based management architecture,
and deployed as VNFs on a cloud platform and then Management Function plays the role of either
network services could be provided by chaining these Management Service producer or Management Service
network functions. A framework is needed to manage consumer. The management system shall be capable to
the life cycle of NFV-based network services. To this consume NFV MANO interface (e.g. Os-Ma-nfvo, Ve-
end, the European Telecommunications Standards Vnfm-em and Ve-Vnfm-vnf reference points).
Institute (ETSI) has presented the ETSI NFV Producer of management services can consume
architecture [19]. A set of SDN applications and management interfaces provided by NFV MANO for
controllers can be added to the standard NFV following purposes:
architecture to support network slicing [20]. Three main
working domains are identified in NFV: (1) Virtualized • Network Service Lifecycle Management (NS
Network Functions, (2) NFV Infrastructure (NFVI), (3) LCM), Performance Management (PM), Fault
NFV Management and Orchestration (MANO). Management (FM), Configuration Management
MANO covers the orchestration and lifecycle (CM) on VNF supporting NS
management of physical and/or software resources that • VNF Lifecycle Management (VNF LCM),
support the infrastructure virtualization, and the Performance Management (PM), Fault
lifecycle management of VNFs. It focuses on all Management (FM), Configuration Management
virtualization-specific management tasks necessary in (CM) on resources supporting VNF
the NFV framework. MANO includes three functional 3Gpp identifies some management functions related
blocks: to network slicing management, each provides some
management services and may consume some
• Virtualized Infrastructure Manager (VIM): management services produced by other functional
comprises the functionalities that are used to blocks [22]:
control and manage the interaction of a VNF with
computing, storage and network resources under • Communication Service Management Function
its authority, as well as their virtualization. (CSMF): this function is responsible for
• NFV Orchestrator (NFVO): is in charge of the translating the communication service-related
orchestration and management of NFV requirement to network slice related requirements.
infrastructure and software resources, and The CSMF communicates with the Network Slice
realizing network services on NFVI. Management Function (NSMF)
• VNF Manager (VNFM): is responsible for VNF • Network Slice Management Function (NSMF):
lifecycle management (e.g. instantiation, update, this function is responsible for the management
query, scaling, and termination). (including lifecycle) of NSIs. It derives network
slice subnet related requirements from the
Network slicing, especially in cases where the network slice related requirements. NSMF
number of network slices is large, increases the communicates with the NSSMF and the CSMF. It
complexity of network management. Therefore, it is provides the management services for one or more
vital to find solutions for automated slice management NSI and may consume some management
and orchestration. To provide the capability of services produced by other functional blocks.
automated and flexible slice management, the concept • Network Slice Subnet Management Function
of Communication Service (CS) introduced by 3GPP, (NSSMF): This function is responsible for the
could be used. A CS is provided by the network management (including lifecycle) of NSSIs. The
operator and is presented to a service operator or a NSSMF communicates with the NSMF. It
business owner. The service operator uses this CS provides the management services for one or more
either for its own communication requirements or to NSSI and may consume some management
provide services to its users. Such a CS requires an E2E services produced by other functional blocks.
network to be run on it. Indeed, a CS uses a Network • Network Function Management Function
Slice Instance (NSI) to provide its service. Different (NFMF): Provides the management services for
aspects of managing network slices is defined by 3GPP managing one or more Network Functions (NF)
in TR28.801 [21]. Complete management of a NSI and may consume some management services
includes managing all its operational functions as well produced by other functional blocks.
as resources needed for supporting its communication • Management Data Analytics Function (MDAF):
services. An NSI may be a combination of network Provides the Management Data Analytics Service
20 Volume 15- Number 4 – 2023 (15 -24)

for one or more NF, NSSI and/or NSI, and may

consume some management services produced by
other functional blocks.
Fig. 8 shows 3GPP Management Functions and a
deployment scenario for NSSI management with
interface to NFV-MANO [22]. The entity denoted as
NSSMF (NSS Management Function), is capable of
consuming the VNF LCM and NS LCM related
services provided by the NFV-MANO (NFVO). Same
entity is also a provider of the NSS related management
services. The entity denoted as NFMF (NF
Management Function), is capable of application level
management of VNFs and PNFs and is a producer of
the NF Provisioning service that includes Configuration
Management (CM), Fault Management (FM) and
Performance Management. Same entity is consumer of
the NF Provisioning service produced by VNFs and Fig. 8. 3GPP Management Functions and a deployment scenario
for NSSI management with interface to NFV-MANO [22]
PNFs.
Using 3GPP slice management functions and
including them in ETSI NFV basic architecture as
described in [7] and considering slice management and
orchestration architecture presented in [17] , we have
introduced an extended architecture for slice
management and orchestration based on NFV MANO
framework which is expected to support network
slicing for different eMBB, URLLC, and mMTC slices.
Fig. 9 shows this architecture that is in accordance with
NGMN conceptual network slicing model [4] where
business layer is added on top. Hence, the layered
architecture consists of 4 layers: (1) Business layer, (2)
Communication Service Instance layer, (3) Network
Slice Instance layer, and (4) Resource layer. The
OSS/BSS domain is enhanced with 3 management
functions including Communication Service
Management Function (CSMF), Network Slice Fig. 9. NFV Based Network Slice Management and Orchestration
architecture
Management function (NSMF), and Network Slice
Subnet Management Function (NSSMF). As shown in
Fig. 10, the Os-Ma reference point can be used for the A. Architectural Functional Blocks description
interaction between 3GPP slicing related management The architectural framework (Fig. 10) specifies
functions and NFV-MANO to manage NSI lifecycle functional blocks and the main reference points
management as well as fault, configuration, accounting, between such blocks. The functional blocks are:
performance, and security management (FCAPS). To Virtualized Network Function (VNF): A VNF is
properly interface with NFV-MANO, the NSMF and/or a virtualization (software-based) of a network function.
NSSMF need to determine the type of Network Examples of NFs are 3GPP 5GC network elements,
Services (NSs) or set of NSs, VNF and PNF that can e.g., AMF, NSSF, UPF, SMF; conventional network
support the resource requirements for a NSI or NSSI, functions, e.g., Dynamic Host Configuration Protocol
and whether new instances of these NSs, VNFs and the (DHCP) servers, firewalls; and RAN layer functionality
connectivity to the PNFs need to be created or existing and protocols s, e.g., NAS, RRC, SDAP, PDCP, RLC,
instances can be re-used. MAC, and also some functionalities of PHY as well as
VI. PROPOSED NETWORK SLICING MEC applications and cache servers.
ARCHITECTURAL (NSA) FRAMEWORK Element Management (EM): The Element
In this section, an SDN/NFV based architecture for Management performs the typical management
5G network slicing is proposed. We have included functionality (fault, configuration, accounting,
3GPP network slicing concepts as well as ETSI NFV performance, and security) for one or several VNFs. An
management and orchestration (MANO) block. This EM itself may be implemented as a VNF. For
architecture provides different functional split options compatibility with 3GPP network slicing concepts, we
for radio access stack protocol [23], [24], [25]. This use NFMF instead of EM.
capability enables us to deploy RAN sub-layers in NFV Infrastructure (NFVI): The NFV
central unit (CU or BBU) or in distributed unit (DU or Infrastructure is the totality of all hardware and
RRH) for any slice type based on its slice-specific software components which build up the environment
requirements. Fig. 10 shows proposed architecture. It in which VNFs are deployed, managed and executed
covers all three network domains including core [19]. In NFV, the physical hardware resources consist
network, transport network, and radio access network. of computing, storage and network that provide
processing, storage and connectivity to VNFs through
Volume 15- Number 4 – 2023 (15 -24) 21

the virtualizati on layer. The virtualization layer control connectivity between different PoPs as well as
abstracts the hardware resources and decouples the backhaul and fronthaul links, SDN controllers at the
VNF software from the underlying hardware, thus position of VIMs are used [20]. Therefore, we have
ensuring a hardware independent lifecycle for the evolved the VIM entity with 6 sub-blocks: Core
VNFs. The use of hypervisors is one of the present Network VIM (CN VIM), Core Network SDN
typical solutions for the deployment of VNFs [26]. Controller (CN SDN-C), Backhaul SDN Controller
When virtualization is used in the network resource (BH SDN-C), RAN Central Unit VIM (CU VIM),
domain, network hardware is abstracted by the Fronthaul SDN Controller (FH SDN-C), and Radio
virtualization layer to realize virtualized network paths Resource Controller (RRC). Note that CU-RAN SDN
that provide connectivity between VMs of a VNF Controller is not considered because different BBUs are
and/or between different VNF instances [27]. RAN not connected to each other and each of their VIMs
virtualization due to limited radio resources and time manages its internal connectivity.
varying channels is more complicated [28]. However,
RAN slicing has been emerged as a solution [29]–[33]. Virtualized Network Function Manager
(VNFM): functional block that is responsible for the
The NFV Infrastructure can span across several lifecycle management of VNF. For this end, it uses
domains and locations. As shown in Fig. 10, there is VNF Descriptor (VNFD). Each PoP Provides a VIM
five different domains in accordance with 5G mobile
network architecture (Fig. 1) including 5G Core Network Functions Virtualization Orchestrator
Virtualization Infrastructure (5GC NFVI), Backhaul (NFVO): Functional block that manages the Network
Link’s vitalization infrastructure (BHLVI), RAN Service (NS) lifecycle based on Network Service
Central Unit Virtualization Infrastructure (CU NFVI), Descriptor (NSD). It coordinates the management of
Fronthaul Link’s vitalization infrastructure (FHLVI), NS lifecycle, VNF lifecycle (supported by the VNFM)
and Distributed Unit RAN slicing Infrastructure and NFVI resources (supported by the VIM) to ensure
(RANSI). The 5GC NFVI and CU NFVI are cloud- an optimized allocation of the necessary resources and
based and deployed in the form of datacenters (DCs). connectivity. It has two set of functions performed by
the Resource Orchestrator (RO) and Network Service
Management and Orchestration (MANO): Orchestrator (NSO).
Performs all the virtualization-specific management,
coordination, and automation tasks in the NFV Communication Service (CS): communication
architecture. service Instances are laid at service instance layer. We
consider a communication service instance as an end-
Virtualized Infrastructure Manager (VIM): to-end NFV composite Network Service Instance that
Functional block that is responsible for controlling and spans over all administrative domains. A CS matches
managing the NFVI compute, storage and network an NSI which may be consisted of one or some NSSI
resources. To manage the resources and their which in turn correspond to an NFV Network Service.
connectivity of a single PoP within its administrative Three different type of communication services i.e.,
domain, the infrastructure provider exposes a eMBB, uRLLC, and mMTC are considered.
Virtualized Infrastructure Manager (VIM). However, to

Fig. 10. Proposed 5G Mobile network slicing architecture

22 Volume 15- Number 4 – 2023 (15 -24)

Oper ation Support System/Business Support dispatches them to the network slice instances in an
System (OSS/BSS): a variety of software systems and optimal way. This optimization means that all the NSIs
management applications used to provide and operate are simultaneously provided with the resources needed
network services. In our view to OSS/BSS, all players to meet their requirements, while preserving their
involved in creation, provision, and operation of a performance isolation. The resource requirements of
communication services such as infrastructure each NSI are stated by its NST.
provider, network slice provider, and network service
provider use their specific OSS/BSS applications at An NSI uses its assigned resources to run instances
different administrative level to perform infrastructure, of VNFs. These VNF instances are chained to build up
network, and service-related management tasks. the required network service instances, following the
However, we didn't show the infrastructure provider specificities given in the NSSMF. At infrastructure
OSS/BSS in Fig. 10. level, VNF instances are executed on virtualization
containers. These virtualization containers are deployed
B. Network slice Creation inside one or more PoPs, according to the geo-location
A complete study on network slice creation phase is requirements of the VNFs.
presented in [34]. Authors introduce Instantiation Level To preserve management isolation across NSIs,
concept for Network Service and Network Slice. They each NSI must have its own management plane. This
also include two functional blocks, Network Slice plane consists of four functional blocks: VNF Manager
Orchestrator and Resource Orchestrator in their (VNFM), NS Orchestrator, Tenant SDN Controller,
architecture. However, we focus on 3GPP management and Network Slice Manager [34]. According to the
services and management functions to provide network one-to-one correspondence between each network slice
slice instances. and the communication service, in Fig. 10, we have
Network slices are created by Network Slice presented a limited level of this management blocks in
(Communication service) provider. The network slice the form of dotted boxes in the communication service
provider uses infrastructure resource to provision provider application. The management plane of CS
network slice instances. To this end, Network Slice consists of four functional blocks: CS NFVM, CS
Provider uses three management functions including NFVO, CS SDN-C, and CS Manager. Indeed, the CS
CSMF, NSMF, and NSSMF and a set of provider could control, manage, and orchestrate its own
Communication Service Templates (CST), Network CS-related NSI, NSSIs, VNFs, NSs, and SDN
Slice templates (NST), and Network Slice Subnet controller in a limited level through APIs exposed by
Templates (NSST) in its OSS/BSS domain. CST, NST, network slice provider.
and NSST are descriptions of the structure (and VII. PROPOSED NSA COMPARISION
contained components) and configurations of CSI, NSI,
and NSSI respectively. The creation and modification In this section, we have compared some famous
of these templates are out of the scope of this work. network-slicing architectures. These architectures are
discussed in the Related Works section. Comparison
Communication Service Customer (CSC) requests criteria are enabling technologies (SDN/NFV), Service
a CS from Communication Service Provider (CSP). Based Architecture (SBA), Network domain
CSMF adapts the requested CS with existing CSTs and (Core/Radio Access), Network Slice Creation
selects the closest CST and customizes it to satisfy CS /Operation Phases, and RAN Split Capability. Results
requirements (such as network type, network capacity, are presented in Table 1. As shown in the last row of
QoS requirements, etc.). Then, based on this CST, Table 1 and discussed in the previous section, our
translates the CS requirement to network slice proposed network slicing architecture is complete. It
requirements. NSMF receives NS requirements from uses NFV management and orchestration capabilities
CSMF and derives network slice subnet(s) and SDN controllers while being compatible with
requirements from network slice requirements using 3GPP Service-Based architecture. It also provides life
NST. NSMF is responsible for management and cycle management of network slices in both creation
orchestration of NSI. NSSMF receives NSS and operation phases in both core and radio access
requirements from NSMF and defines NFV network domains of the 5G network. In addition, we have
service(s) requirements from network slice subnet included a functional RAN layer split to lay the
requirements using NSSTs and send it to NFV-MANO corresponding layers in centralized or distributed units
to deploy NSSIs. NSSMF is responsible for according to the requirements of each eMBB, mMTC,
management and orchestration of NSSI. or URLLC slice. Performance evaluation of the
C. Operation Phase presented model in different domains, including
Network Slice Management and Orchestration,
NSMF and NSSMF use NFV-MANO management Admission control, RAN Resource Allocation, and 5G
services to manage and orchestrate available resources RAN Functional Layer Split based on this model, are
between network slice instances and network slice the subject of our current studies and will be published
subnet instances, respectively. The NFVO as a soon.
Resource Orchestrator uses the finite set of resources
that are exposed by the underlying VIMs/SDN-Cs and
Volume 15- Number 4 – 2023 (15 -24) 23

TABLE I. NETWORK SLICING ARCHITECTURES COMPARISON

NS Architecture SDN NFV SBA RAN Split CN/RAN NSI Creation/Operation
ONF [6] Yes No No No CN Operation
ETSI NFV [7] No Yes No No CN Creation/Operation
NGMN [13] No Yes No Yes CN/RAN Operation
5GNORMA [2] Yes No No Yes CN/RAN Operation
5GPPP [14] No Yes Yes No CN/RAN Creation/Operation
3GPP SA [15] No Yes Yes No CN/RAN Creation/Operation
Proposed NSA Yes Yes Yes Yes CN/RAN Creation/Operation

VIII. FUTURE WORKS functions for network slicing were introduced, and
based on this we presented an NFV based slice
This presented architecture is an overall 5G mobile management and orchestration architecture using 3GPP
network slicing framework. It provides network slicing CSMF, NSMF and NSSMF management functions.
capability in both creation and operation phases. Based Finally, an SDN/NFV based architecture for 5G
on this framework, the following works are being done: network slicing is proposed. We have included 3GPP
• Multi SLA Network Slice Management and network slicing concepts as well as ETSI NFV
Orchestration that is a network slice creation management and orchestration (MANO) block in
algorithm based on tenant’s SLA. combination with SDN controllers. Our proposed
• Multi-Level Admission control for 5G Network architecture provides different functional split options
Slicing that is a two-level admission control for radio access stack protocol. This capability enables
algorithm in operation (run time) phase. us to deploy RAN sub-layers in central unit (CU or
• Hierarchical OMA/NOMA RAN Resource BBU) or in distributed unit (DU or RRH) for any slice
Allocation for 5G Network Slicing. In this work a type based on its slice-specific requirements. It covers
two-level resource allocation algorithm is being all three network domains including core network,
presented. At the first level (Slice Provider level), transport network, and radio access network. In the
resources are pre-allocated to tenants based on following, we explained the operational blocks of the
their SLAs. Then in the second level (Slice proposed architecture and showed how this architecture
Operator level), slice’s pre-allocated resources are can create and operate various MBB, mMTC and
allocated to users based on the user priority. uRLLC slices in the core and radio access sections
• 5G RAN Functional Layer Split to serve eMBB, using SDN controllers, virtualization, NFV orchestrator
uRLLC, and mMTC slices. Based on the RAN and 3GPP management functions.
functional layer split capability of the proposed REFERENCES
framework, we are working on a model to deploy
[1] International Telecommunication Union, “Minimum
RAN sub-layers in central unit (CU or BBU) or in requirements related to technical performance for IMT-2020
distributed unit (DU or RRH) for any slice type radio interface(s),” ITU-R M.2410-0, vol. 0, p. 9, 2017.
based on its slice-specific requirements [2] 5GNORMA, “5G NORMA Network Architecture Final
report,” 5G NORMA Proj., pp. 1–159, 2017.
IX. CONCLUSION [3] J. Ordonez-Lucena, P. Ameigeiras, Di. Lopez, J. J. Ramos-
Network slicing is a basic requirement in new Munoz, J. Lorca, and J. Folgueira, “Network Slicing for 5G
mobile networks. Although many steps have been taken with SDN/NFV: Concepts, Architectures, and Challenges,”
IEEE Commun. Mag., vol. 55, no. 5, pp. 80–87, 2017, doi:
in this regard, and some frameworks have been 10.1109/MCOM.2017.1600935.
proposed for network slicing, but as far as we have [4] NGMN Alliance, “Description of Network Slicing Concept,”
studied, there is a lack of general architectural ngmn, vol. 1, no. 1, p. 7, 2016.
framework both about enabling technologies and [5] 5GPPP, “View on 5G Architecture V4,” 5G-PPP Initiat., no.
diversity of provided services, as well as compatibility October, 2021, doi: 10.5281/zenodo.5155657.
with the 3GPP 5G system architecture. [6] Open Networking Foundation, “Applying SDN Architecture to
5G Slicing,” ONF TR-526, no. 1, pp. 1–19, 2016.
In this paper, we have introduced the concept of [7] ETSI, “Network Functions Virtualisation ( NFV ) Release 3 ;
network slicing as well as enabling technologies Evolution and Ecosystem ; Report on Network Slicing Support
including SDN and NFV. We also reviewed 3GPP 5G with ETSI NFV Architecture Framework,” ETSI GR NFV-EVE
service-based architecture and present two 012 V3.1.1, vol. 1, pp. 1–35, 2017.
representation of this architecture, namely Service- [8] L. U. Khan, I. Yaqoob, N. H. Tran, Z. Han, and C. S. Hong,
Based representation and Reference-Point “Network Slicing: Recent Advances, Taxonomy,
Requirements, and Open Research Challenges,” IEEE Access,
representation, as well as description of 5G network vol. 8, pp. 36009–36028, 2020, doi:
functions. In addition, 3GPP 5G NR is introduced along 10.1109/ACCESS.2020.2975072.
with description of NG interface and 5G NR protocol [9] A. A. Barakabitze, A. Ahmad, R. Mijumbi, and A. Hines, “5G
stacks. Next, we expressed the concept of 5G mobile network slicing using SDN and NFV: A survey of taxonomy,
network slicing and presented a 5G network slicing architectures and future challenges,” Comput. Networks, vol.
architecture based on 3GPP 5G system architecture to 167, p. 106984, 2020, doi: 10.1016/j.comnet.2019.106984.
support various eMBB, mMTC and uRLLC slices [10] P. Subedi et al., “Network slicing: a next generation 5G
while separating the radio access network and core perspective,” Eurasip J. Wirel. Commun. Netw., vol. 2021, no.
1, 2021, doi: 10.1186/s13638-021-01983-7.
network segments focusing on shared network
[11] D. Alotaibi, “Survey on Network Slice Isolation in 5G
functions and dedicated network functions for each Networks: Fundamental Challenges,” Procedia Comput. Sci.,
slice. Then the network slice lifecycle management was vol. 182, pp. 38–45, 2021, doi: 10.1016/j.procs.2021.02.006.
explained and 3GPP management and orchestration
24 Volume 15- Number 4 – 2023 (15 -24)

[12] C. Ssengonzi, O. P. Kogeda, and T. O. Olwal, “A survey of

deep reinforcement learning application in 5G and beyond [32] R. Ferrús, O. Sallent, J. Pérez-Romero, and R. Agustí, “On 5G
network slicing and virtualization,” Array, vol. 14, p. 100142, Radio Access Network Slicing: Radio Interface Protocol
Jul. 2022, doi: 10.1016/j.array.2022.100142. Features and Configuration,” IEEE Commun. Mag., vol. 56,
[13] NGMN Alliance, “NGMN 5G White Paper,” ngmn, pp. 1–125, no. 5, pp. 184–192, 2018, doi: 10.1109/MCOM.2017.1700268.
2015. [33] P. Popovski, K. F. Trillingsgaard, O. Simeone, and G. Durisi,
[14] 5GPPP, “View on 5G Architecture V2,” 5G-PPP Initiat., 2017, “5G wireless network slicing for eMBB, URLLC, and mMTC:
doi: 10.13140/RG.2.1.3815.7049. A communication-theoretic view,” IEEE Access, 2018, doi:
[15] 3GPP, “5G System Architecture for the 5G System,” 2019. 10.1109/ACCESS.2018.2872781.
[Online]. Available: [34] J. Ordonez-Lucena et al., “The Creation Phase in Network
https://www.etsi.org/deliver/etsi_ts/123500_123599/123501/1 Slicing: From a Service Order to an Operative Network Slice,”
5.09.00_60/ts_123501v150900p.pdf. 2018 Eur. Conf. Networks Commun. EuCNC 2018, pp. 31–36,
[16] 3GPP, “5G NR Overall description,” 2018. [Online]. 2018, doi: 10.1109/EuCNC.2018.8443255.
Available:
https://portal.etsi.org/TB/ETSIDeliverableStatus.aspx.
[17] S. Zhang, “An Overview of Network Slicing for 5G,” IEEE Mohamad Ali Bayat
Wirel. Commun., vol. 26, no. 3, pp. 111–117, 2019, doi: Mokhtari received his B.Sc.
10.1109/MWC.2019.1800234. and M.Sc. degrees in
[18] 3GPP, “5G; Management and orchestration; Concepts, use Electronics Engineering from
cases and requirements,” ETSI TS 128 530, vol. 0, pp. 0–32, Ferdowsi University of
2020.
Mashhad, Mashhad, Iran in
[19] ETSI, “Network Functions Virtualisation (NFV); Architectural
Framework,” ETSI GS NFV 002 V1.2.1, vol. 1, no. 12, pp. 1–
1997, and 2001 respectively.
21, 2014, doi: DGS/NFV-0011. He is a Ph.D. candidate of
[20] ETSI, “Network Functions Virtualisation (NFV); Ecosystem; Computer Systems
Report on SDN Usage in NFV Architectural Framework,” Architecture at Tehran
ETSI GS NFV-EVE 005 V1.1.1, vol. 1, no. 12, pp. 1–125, 2015. University. His research interests include Next
[21] 3GPP, “Study on Management and Orchestration of Network Generation Mobile Networks, Network Function
Slicing for Next Generation Network,” 3GPP TR 28.801, 2018. Virtualization, Network Slicing, and Software-Defined
[22] 3GPP, “5G; Management and orchestration; Architecture Networks.
framework,” 2018.
[23] NGMN Alliance, “NGMN Overview on 5G RAN Functional Nasser Yazdani received the
Decomposition,” ngmn, vol. v1.0, 2018. B.Sc. degree in Computer
[24] NGMN Alliance, “5G RAN CU - DU Network Architecture , Engineering from the Sharif
Transport Options and Dimensioning,” ngmn, vol. v1.0, pp. 0– University of Technology,
35, 2019. Tehran, Iran, in 1985, and the
[25] 5GNORMA, “RAN architecture components final report,” 5G Ph.D. degree in Computer
NORMA Proj., 2017.
Science and Engineering from
[26] ETSI, “Network Functions Virtualisation (NFV);
Virtualisation Technologies; Hypervisor Domain
Case Western Reserve
Requirements specification; Release 3,” ETSI GS NFV-EVE University, Cleveland, OH,
001 V3.1.1, vol. 1, no. 7, pp. 1–15, 2017. USA, in 1997. He worked at the
[27] A. Fischer, J. F. Botero, M. T. Beck, H. de Meer, and X. Iran Telecommunication Research Center (ITRC),
Hesselbach, “Virtual Network Embedding: A Survey,” IEEE Tehran, as a Consultant, a Researcher, and a Developer
Commun. Surv. Tutorials, vol. 15, no. 4, pp. 1888–1906, 2013, for few years. He later worked at different companies
doi: 10.1109/SURV.2013.013013.00155.
and research institutes in USA. He joined the
[28] C. Liang and R. Yu, “Wireless Network Virtualization: A Department of Electrical and Computer Engineering,
Survey, Some Research Issues and Challenges,” IEEE
Commun. Surv. Tutorials, no. c, pp. 1–1, 2014, doi: University of Tehran, Tehran, in September 2000. He
10.1109/COMST.2014.2352118. then initiated different research projects and labs in
[29] O. Sallent, J. Perez-Romero, R. Ferrus, and R. Agusti, “On high-speed networking and systems. He is currently a
Radio Access Network Slicing from a Radio Resource Full Professor with the School of Electrical and
Management Perspective,” IEEE Wirel. Commun., vol. 24, no. Computer Engineering, University of Tehran. His
5, pp. 166–174, 2017, doi: 10.1109/MWC.2017.1600220WC. research interests include Networking, Packet
[30] D. Marabissi and R. Fantacci, “Highly Flexible RAN Slicing Switching, Access Methods, Operating Systems and
Approach to Manage Isolation, Priority, Efficiency,” IEEE
Access, vol. 7, pp. 97130–97142, 2019, doi:
Database Systems.
10.1109/ACCESS.2019.2929732.
[31] P. L. Vo, M. N. H. Nguyen, T. A. Le, and N. H. Tran, “Slicing
the edge: Resource allocation for RAN network slicing,” IEEE
Wirel. Commun. Lett., vol. 7, no. 6, pp. 970–973, 2018, doi:
10.1109/LWC.2018.2842189.