SRM INSTITUTE OF SCIENCE AND TECHNOLOGY

RAMAPURAM
FACULTY OF ENGINEERING & TECHNOLOGY

DEPARTMENT OF ELECTRONICS AND COMPUTER ENGINEERING

Unit 2: 5G Network Architecture

Introduction to 5G Network Architecture; 5G Core Network (5GC); Radio Access Network; Network
Slicing: Concept and Implementation; Virtualization in 5G; Software-Defined Networking (SDN) in
5G; Edge Computing; Mobile Edge Computing; Quality of Service requirements.

Introduction to 5G Network Architecture

• 5G represents a paradigm shift for cellular/broadband networks. It literally pushes toward
virtualization and away from the implementation of monolithic network entities.
• It brings a number of challenges and unknowns that the telecom sector has not so far embraced.
The transformation from 3G/4G to 5G will be profound and challenging as users are creating
demand requiring new approaches for connectivity, bandwidth, and network architecture.
• This section will articulate three architectural views from a very high level to a much deeper
degree for the understanding of the readers.
• 3GPP in collaboration with its seven organizational partners* (or SDOs) is working to produce
5G specifications which will include network architecture as well. This work is expected to be
completed in three releases, that is, Rel-14, Rel-15, and Rel-16.
• 3GPP is planning to submit such specifications (proposals) as per ITU-R timelines. Proposals are
also expected from IEEE-SA and others.

30,000 Ft Bird’s Eye View

• The design of 5G network is still evolving and the sector will take some time to thrash out the
details.
• Nevertheless, it is expected that the architecture will utilize the cloud functionality and small
cells to the deepest extent. A 30,000 ft bird’s eye view, shown in Figure 8.4, highlights the key
segments of the architecture.
• It shows that small cells are connected with either view optical fiber cable or radio waves to the
cloud radio access network (BBU pool). The pools of BBU are connected (which is called the
fronthaul) to the RRUs (small cells or macro cells) and are backhauled to the core network.
• 5G core network in the form of cloud is integrated with existing 3G/4G core networks and
provides connectivity to the Internet.

Figure.2.1.30,000 ft 5G network view

High Level 5G Network View
• 5G networks will consist of thousands of small cells making networks ultradense. This network
densification is required to meet the latency and throughput requirements.
• Small cells are expected to carry a major portion of traffic with overall data volume expected to
grow exponentially as predicted by many studies.
• Bringing small cells closer to the user will reduce latency and increase overall network efficiency
by creating subnetworks.
• These subnetworks can have the functionality to route data traffic locally for the local users’ calls
who are communicating with each other while sending the signaling to the main network as
shown in Figure 2.2.
• The net result from local switching is in the savings of network resources and cheaper data plans
for the customers.

FIGURE.2.2. Local routing.

• C-RAN will bring cloud functionality and likely cognitive capability to the RAN. In such flexible
RAN, RRUs do not need to stick to a particular BBU, but instead they can talk to any of the BBUs
of a specific pool.
• Cloud functionality will probably not exist in transmission networks including backhaul and will
continue to rely on fiber and radios waves-based systems. Additionally, DSL may continue for
connectivity with indoor small cells.
• 5G networks are expected to integrate NFV and SDN along with separate user and control planes.
Nevertheless, at least for now, four areas of cellular networks, namely radio access, transport,
core networks and operations, will continue to exist in 5G architectures as shown in Figure 2.3.

FIGURE 2.3. High level 5G network architecture.

 • A major challenge for 5G radio access networks is the requirement for efficient integration of an
additional layer of small cells with the existing macro cell network.
• The continued technological upgrades and growth in the number of macro cells will further
augment this challenge. In the case of 5G, small cells are deployed as light RRUs along with macro
RRUs.
• The light RRU can be installed on streetlight poles and underground subways whereas macro
RRU is suitable for traditional BTS towers. Both types of RRUs will connect to pools of BBUs
through a fronthaul interface.
• Fronthaul is the new area in the transport network which connects the radio heads to the
centralized pool of BBUs through fiber.
• The current choices for fronthaul interfaces are CPRI, ORI (Open Radio equipment Interface) or
OBSAI (Open Base Station Architecture Initiative). However, these may encounter capacity
bottlenecks and thus require the players to look for alternatives.
• The BBUs’ pools are connected to the core network through a combination of wired and wireless
technologies. These include fixed links such as dedicated fiber, together with wireless options
such as microwave radios and radio over fiber.
• The core network will be connected to Internet and to connect different metropolitan cities,
existing long haul WDM (wave division multiplexing) connectivity can be upgraded.
• The core network can be split into control and user plane domains where a combined control
entity (CCE) integrates the control functions of the 5G core network with the core functions of
MME, SGW-C, and PGW-C.
• Similarly, a combined data gateway (CDG) performs data forwarding functions of 5G core
network, SGW-D, and PGW-D.

5G Network Architecture (In-Depth View)

• The 5G mobile network will include both physical and virtual functions and cloud deployments.
It is also clear that 5G needs to account for existing 3G/LTE/LTE-Advanced network
deployments.
• An in-depth view of 5G network architecture is shown in Figure 2.4. At a high level, two options
have been envisioned for RATs.
• One difficult possibility is having a single unified RAT that is optimized for different frequencies,
various use cases (IoT, V2X [vehicle-to-everything], etc.) and a variety of services.
• Designing such a RAT is highly challenging and costly, requiring refarming of the existing
spectrum used for legacy LTE and LTE-Advanced technologies.
• Spectrum refarming may not be ideal in many cases, thus a better approach may be 5G carrier
aggregation with LTE/LTE-Advanced carriers during the initial period of 5G rollout.
• Alternately, multiple RATs could complement each other where one using a high frequency band
may provide capacity and high data rates in dense urban areas, whereas another RAT operating
in a lower frequency band may be used for IoT and in rural areas for extended coverage.
 FIGURE 2.4. 5G network architecture (in-depth view).
• The transport network has to be flexible and dynamic to address the needs of future networks.
Fronthaul, backhaul, metro, long distance, and international connectivity have to be insured by
the transport network.
• Most of the connectivity will be provided through optical fiber cable, but some will require radio
waves-based systems.
• 3GPP envisions a logical split between RAN and core network to allow independent evolution of
either one. Furthermore, a split of control and user planes is foreseen for 5G networks.
• In RAN, for example, it will allow macro cells to handle the control plane while small cells
(particularly in the context of millimeter waves) handle the user plane. However, the exact
nature of such a split is still under investigation.
• A typical split could be between physical and MAC layers as shown in Figure 8.7 where the
fronthaul is primarily supported by fiber.
• In this option, RRU supports modulation and precoding functions of the physical layer, while
BBUs will support MAC, RLC, PDCP and RRC, and FEC and HARQ (Hybrid Automatic Repeat
Request) functions of the physical layer.
• This is a centralized approach without the need of a high bandwidth support in the fronthaul.
• NFV and SDN are essential in 5G networks to reduce costs and bring added value to network
infrastructure.
• NFV is the process of moving/forwarding tasks such as load balancing, firewalls, and so on away
from dedicated hardware into a virtualized environment.
• NFV enables the execution of software-based network function on general purpose hardware by
leveraging virtualization techniques.
• The virtualization technologies allow breakup of the software of network functions from
dedicated hardware.
 • Softwarization allows implementation of network functions in software, including virtualization
of such functions and programmability by setting appropriate interfaces. It is an approach to use
software programming to design, implement, and maintain network equipment and services. In
SDN, the control plane is decoupled from the data plane and is managed by a logically centralized
controller that has a holistic view of the network.
• Softwarization in RAN may allow some functions such as PDCP and RRC to be implemented as
VNFs. Softwarization can also be used to implement certain core and transport functions.
• The original aim of combining NFV and SDN was to decouple services from physical resources
allowing flexibility and adaptability in the network. When NFV and SDN come together, they
provide the additional benefit of detaching lifecycle management from physical constraints.
• Network slice supports the connectivity of a particular use case through a collection of 5G
network functions, and specific configurations in RAN, transport, and core networks. Network
functions provide connectivity, storage, and computation.
• Details on network functions can be found in. Finally, 5G is not all about connectivity, but also
demands high end computation and storage. Computation and storage requirements also vary
among the different network areas and elements.
• For example, a BBU pool may have less stringent needs as compared to a core network packet
gateway. Similarly, the transport network encompasses several aggregation nodes that need to
offer computing and storage capabilities.

5G Core Network (5GC)

• Core networks traditionally have been designed as a single architecture addressing a range of
requirements and supporting backward compatibility.
• This one size fits all approach has been successful in keeping the costs down to a reasonable
level and by supporting legacy circuit switched and today’s packet switched functionalities.
• This core network, however, is rigid in the sense that it is not flexible enough to accommodate
the customized and variable connectivity needs of individual users and businesses that are
expected in the future.
• However, with virtualization, NFV, SDN, and network slicing, it is possible to make core networks
more flexible and scalable. Thus, the next generation core network is expected to exist in a cloud-
based environment with a high degree of virtualization and software-based networking.
• Such flexibility is needed to support a variety of access networks such as 3G, LTE, 4G, WiFi, and
tomorrow’s 5G.

Components of Core Network/High Level Architecture

• The current EPC will further evolve to support virtualization and network slicing to become NGC
applicable for 5G networks.
• Network slicing is often termed as logical instantiation of a network possibly due to
virtualization technologies. The concept is seen as the natural extension/evolution of the
current network sharing methodologies.
• Network slicing is one of the promising techniques that will likely exist in both radio access and
core networks. It allows multiple logical networks to be created on top of a common physical
infrastructure.
• Either DCN (Dedicated Core Network) or a combination of NFV an SDN can be used as a
technology to enable network slicing along with orchestration and analytics. DCN or Decor as
defined in 3GPP TS 23.401 is a feature that enables an operator to deploy multiple logical mobile
core networks connected to the same RAT or multiple RATs (e.g. GERAN, UTRAN, E-UTRAN, WB-
E-UTRAN and NB-IoT).
 • A DCN consists of one or more MME/SGSN and it may be comprised of one or more
SGW/PGW/PCRF. This feature enables subscribers to be allocated to and served by a DCN based
on subscription information (e.g., “UE Usage Type”).
• With 5G, a single terminal can use multiple services with different characteristics almost
simultaneously. In such cases, a network slice can be created for each service, requiring all such
slices to coordinate control for that particular single terminal.
• These slices can be mapped to respective radio and core network slices to provide end-to-end
connectivity. The methodology is currently being specified for selecting radio/core networks
particularities for supporting slicing in existing as well as in future 5G systems.
• Control and User Planes’ Separation: The separation of control and user planes is one of the key
principles of 5G core network architecture.
• 3GPP started a study in TR 23.714 [28] on user/ control planes’ separation involving core
network elements such as P-GW, Traffic Detection Function (TDF), and so on. This separation
allows independent scaling of each plane and migration toward cloud-based architecture.
• For example, the control plane can be placed in a centralized location with complex hardware
and processing capabilities. On the other hand, the user plane can be distributed to a larger
number of local sites making reachability from the perspective of a user easier.
• A good example of this will be content caching in local sites instead of securing it from the main
server sitting thousands of miles away. This separation is the fundamental concept of SDN and
having SDN will make core networks more flexible.

Radio Access Network

• The RAN, also known as the UMTS Radio Access Network is the equivalent of the previous Base
Station Subsystem (BSS) in GSM.
• RAN includes the Node B function and the Radio Network Controller (RNC) function. The Node
B function provides the air interface. The RNC manages the air interface for the overall
network.
• The Radio Access Network (RAN) architecture has evolved across the different generations of
the wireless network, to support the bandwidth and scalability requirements.
• RAN has two distinct units the Remote Radio Head (RRH) and the Baseband Unit (BBU).
• One end of the RRH is connected to the antenna and the other end to the BBU.
• RRH acts as a transceiver converting the analog signals to digital signals and vice versa.

TABLE 2.1 - EVOLUTION OF RAN

• In addition, RRH also does filtering of noise and amplification of signals. The Baseband Unit
(BBU) provides switching, traffic management, timing, baseband processing, and radio
interfacing functions. BBU is typically connected to the RRH using a Fiber link.
• In the traditional 2.5G/3G networks, both the RRH and BBU functions remained in the cell site,
as part of the Base Transceiver Station (BTS).
• In the 4G network, BBU function was moved out of the cell site to a centralized location. BBU
function in a 4G network is hosted in the Central Office and is called as Centralized RAN.
• 4G architecture optionally supports the virtualization of BBUs and when the BBU function is
virtualized, it is also called as Cloud RAN or Virtualized RAN. In a 5G network, virtualization of
BBUs almost becomes mandatory as it helps the service providers to scale the network to
support the various use cases.

FIGURE 2.5 - TRADITIONAL RAN

FIGURE 2.6 - CENTRALIZED RAN

 FIGURE 2.7 - VIRTUALIZED RAN

Network Slicing:
Concept and Implementation
• Network Slicing allows 5G service providers to divide a single physical network (all the
way from the radio to the core network) into multiple virtual networks.
• Each network slice can have different speed limits, different latencies and different
quality of service configuration. Slicing Network is an end-to-end feature offered by 5G
infrastructure ranging from the Radio Access Network (RAN) to the 5G NG-Core.
• Each of the network slices will have its own configuration settings and performance
characteristics. Each network slice is optimised to meet the needs of a given 5G use case.
• For example, smart meters will operate in a separate network slice when compared to
autonomous vehicles. Network Slicing is made possible by the advancements in Network
Functions Virtualization, Software Defined Networks and Cloud.
• A network slicing implementation in the radio network as well as the core network, can
be based on physical resources or virtualized / logical network resources. In the core
network, a network slice can have its own dedicated virtual network functions
instances running on the telco cloud.
• This allows the services providers to offer customized services to its customers and at the
same time optimize the infrastructure costs.

Figure: 2.8. Network Slicing Architecture

• Each network slice provides a set of network capabilities, performance levels and
specific Service Level Agreements (SLAs), to the services that run on top of the
network.
• Services get mapped to network slice instances based on their capabilities,
performance levels and SLAs. A network slice can be dedicated for a given service or
may get shared across multiple services.
• For example, a network slice that supports online gaming service (high bandwidth and
latency sensitive) can also support virtual reality service.
• A network slice that supports web browsing (low bandwidth and latency insensitive
service) can also support IOT services that gather analytics.
• The mapping of RAN slice to the NG-Core Slice is done by the Slice Pairing function. The
slice pairing function can reside in a network management system or as an application
running on top of a SDN Controller.

Requirements for Network Slicing:

The following are some of the requirements for the network slicing implementation in a 5G
network:
• Service provider should be able to configure / manage a network slice dynamically based
on the customer needs.
• Service provides should be able to manage each network slice separately without
impacting the performance characteristics of other network slices.
• Provide security for the services that run on top of a network slice, including
protection of the data that gets transferred over the network slice.
• Service providers should be able to expose Application Programmers Interfaces
(APIs) for its partners, vendors or customers to create and manage network slices.
• Support end-to-end resource management from RAN all the way to the 5G NG-Core.
• The new 5G Core architecture also allows one single device to connect to more than one
slice simultaneously, a feature which was not supported in the EPC architecture defined
for 4G.

Figure: 2.9. Network Slice

• A specific network slice is identified by a parameter called S-NSSAI, short for “Single
Network Slice Selection Assistance Information” and consisting of two sub parameters
the Slice/Service Type (SST) and the optional Slice Differentiator (SD).
• SD is used to differentiate between multiple slices of the same type, hence having the
same SST. The radio network serving the device will use one or more S-NSSAI (SINGLE-
 Network Slice Selection Assistance) values requested by the device to do the initial
selection of AMF (Access and Mobility Management Function).

• The selected AMF will either decide to serve the specific device or make a new slice
selection itself, or it may use the Network Slice Selection Function (NSSF) for this. The
NSSF has as its single role to support the selection of network slices based on a
combination of S-NSSAI values defined for the network, requested by the device and
allowed in the subscription.

Work Slice Management:

• Network Slicing architecture provides mechanisms for service providers to
manage the end-to-end network slice infrastructure both at the Radio Access
Network (RAN) and the Core Network (5G NG-Core).

• The RAN in-turn may be divided into control plane and user plane. The
configuration policies for a network slice will provision and activate services in
both the control and user planes.

There are three distinct layers to be managed for a network slice:

• Service instance layer - This layer consists of instances of services that are
either exposed to customers or business partners of the service provider. For
example, IOT services, Video streaming services and AR/VR services. The
services can be created/managed by a network operator or a third party
service provider.

• Network Slice instance layer – This layer consists of both the RAN slice instance
and the Core Network slice instance. This layer provides the network
characteristics required by a service instance. A network slice instance may get
shared across one or more service instances.

• Resource layer - This layer consists of the actual physical or virtual network
functions that are used for creating a network slice. There are scenarios where
the resources for a network slice may span across multiple operator domains

Figure: 2.10. Network Slicing Management

• Network slice life cycle management function would in-turn interact with the
NFV Management and Orchestrator (MANO) function for managing the
Virtualized Network Functions (VNFs) and would deal with other network
managers that manage Physical Network Function (PNF).

MANO function consists of

o The NFV Orchestrator
o VNF Manager and
o Virtual Infrastructure Manager (VIM).
• Automation of the end-to-end network slice management is very critical to
improve the user experience and operational cost reduction. It can be achieved by
 deploying a SDN Controller in the network. The SDN Controller exposes APIs for
service providers to develop applications that can manage network slices in a
wireless network.
Benefits of Network Slicing:
Network slicing provides a number of benefits to both service providers and customers.
Some of the benefits are
• Reduces operational costs in managing / running the wireless networks, as the 5G
network slicing exposes APIs for programmatically managing the network
infrastructure.
• Today, Mobile Virtual Network Operator (MVNO) services (letting other wireless
service providers share the network infrastructure) require complex manual pre-
provisioning of network infrastructure.
• Network Slicing allows Service Providers to dynamically create, configure and
manage MVNO services.
• Allows Service Providers to offer differentiated services to customers using the
same network infrastructure, without impacting the performance of the services
offered to other customers. For example, support Autonomous Cars and Smart Utility
Meters on the same network infrastructure.
• Allows service providers to monetise the network infrastructure - not just based
on bandwidth consumed, but also based on other parameters such as latency, quality
of service, energy consumption and number of connections

Virtualization in 5G
Virtualization is set to play a major role in the evolution of the fifth-generation (5G) core network.
According to industry experts, 5G will use softwarecentric networking technologies such as
softwaredefined networking (SDN) and network functions virtualization (NFV), and will be natively
cloud based. If correct, this will represent a major transition in system architecture and will require
much greater collaboration across the networking ecosystem. The push to incorporate more cloud- or
software-based components is driven by the need for greater flexibility and scalability to respond to
the demands of radio access technologies that offer more bandwidth, reduced latency, and stringent
quality of service (QoS) requirements. The new 5G core network must be adaptable and better
equipped to handle various devices and manage capacity in near-real time. Mobile network operators
see the advances in cost and efficiency that virtualization brings to other market segments and will use
the emerging 5G technology to determine if these same advances can benefit the mobile market.
SDN and NFV are two key architecture concepts in development to support the flexibility and mobility
demands of the 5G network infrastructure. Virtualizing network functions that were previously
implemented in hardware will allow providers to introduce new features and integrate new standards
at a faster rate. SDN/NFV provides an avenue for providers to decentralize their networks, thereby
increasing flexibility and reducing latency. Two areas where SDN/NFV will benefit 5G networks, and in
some cases even fourth generation (4G) mobile technology, are network slicing and cloud-radio access
network (C-RAN).

Network slicing
Network slicing would promote end-to-end mobile network virtualization by “slicing” the network into
virtual channels. These virtual channels would be autonomous and encompass a set of resources—
physical or virtual—including bandwidth on a network link, processing capacity of servers, processing
capacity of network elements, as well as operations support system (OSS) and business support system
(BSS) processes. Operators could then use these channels to dynamically devote the appropriate
network resources to create a “lane” in the network specifically designed for a particular use or service.
This would accommodate the many use cases being put forth for 5G. The operator-led Next Generation
Mobile Networks (NGMN) Alliance has sought to define categories of 5G use cases (i.e., service types)
that have distinct performance characteristics and commercial potential. In a 2015 white paper, the
NGMN listed eight application categories for 5G.
1. Broadband access in dense areas,
2. Broadband access everywhere,
3. Higher user mobility,
4. Massive Internet of Things,
5. Extreme real-time communications,
6. Lifeline communications,
7. Ultra-reliable communications, and
8. Broadcast-like services.
Each of these service types demand different network requirements that are determined by the types
of traffic being sent and even the types of devices sending the traffic. For example, someone
downloading cat videos will not have the same bandwidth or low latency requirements as a doctor in
Los Angeles performing surgery virtually on a patient in Mumbai. The end-to-end notion of network
slicing could be key to 5G’s ability to effectively accommodate all of these disparate use cases.

Figure.2.11. 5G’s incorporation of SDN and NFV would allow network slices to be created
dynamically and deployed as needed to accommodate a variety of scenarios
The Open Networking Foundation, in an April 2016 white paper, called out SDN’s ability to support
multiple services over a common architecture as a key enabler for network slicing. SDN also allows for
slices to be created dynamically and deployed as needed to accommodate a variety of scenarios (see
figure.2.11). Currently, 4G mobile networks prioritize traffic to get a similar effect, but with more
rigidity and limitations. However, as 4G networks incorporate SDN and NFV, network slicing will
become an option.
C-RAN
There are more than six million base stations (excluding small cells) deployed worldwide across
approximately five million different cell sites serving close to four billion users. The surge in demand
for connectivity has network operators searching for ways to shrink their network footprint, lower
operational expenditures (OPEX), and still meet users’ demand for access. C-RAN meets these
requirements and has either been implemented or trialed by several operators including Verizon, AT&T,
KT (South Korea), and China Mobile. Radio base stations currently depend on special purpose-built
hardware deployed at the cell site. The baseband processing unit (BBU) is the part of the RAN that is
responsible for managing the radio functions (or all functions that require an antenna). The BBU is one
 of the parts of the RAN that can be moved to a central location, creating a pool of BBUs to serve multiple
base stations. C-RAN aims to centralize and virtualize baseband processing to reduce cell site costs and
enable coordinated scheduling of resource blocks across a coverage area [6]. Figure 2 illustrates the
evolution from the classic RAN model to a C-RAN setup. On the left, the classic model has the BBU
deployed at the cell site connecting to the core network over IP/Ethernet transport. Any coordination
between cell sites takes place over the X2 interface, which allows two sites to communicate. In a C-RAN
architecture, the BBUs are pooled at a location away from the cell site. Pooling BBUs negates the need
for the X2 interface as communications between cell sites now takes place internally. This is one of the
reasons for the increased performance in C-RANs. The C-RAN model also makes updating the waveform
and protocols easier as it only requires a software upgrade at the centralized BBU and not at each
individual cell site The move to a C-RAN architecture increases the flexibility of the network by allowing
providers to, in theory, even change the types of RANs used—from 3G to 4G. For instance, in an area
that has a mix of 3G and 4G users, operators can rebalance radio frequency resources by shifting more
resources to 4G when 4G users are prevalent in the covered area. This shuffling of resources ensures
that there is sufficient capacity for subscribers. C-RAN will also enable RAN-as-aService (RaaS), which
will allow operators to rent RAN capacity to other operators.

Figure.2.12. . Unlike the traditional RAN model (on the left), C-RAN technology (on the right)
pools the BBUs at a location away from the cell site offering an increase in performance and
simpler upgrade path
Benefits
• Reduced CAPEX and OPEX.
• Faster service deployment.
• Better resource allocation.
• Improved network efficiency and flexibility.
Challenges & Limitations
• Security risks in shared environments.
• Complexity of managing virtualized resources.
• Interoperability and standardization issues.

Software-Defined Networking (SDN) in 5G

Software-Defined Networking (SDN) is a modern networking approach that separates the
network’s control plane from the data plane.
• Control Plane: Decides how packets should flow through the network.
• Data Plane: Actually forwards the packets.

This separation allows centralized, programmable control over the entire network.

5G networks have complex and diverse requirements — high bandwidth, ultra-low latency, massive
device connectivity, and flexible resource management. SDN helps by:
• Centralizing network control: A logically centralized SDN controller manages and optimizes
resources dynamically.
• Enabling network slicing: SDN supports creating multiple virtual networks (slices) over the same
physical infrastructure — each optimized for different services (e.g., IoT, autonomous cars, AR/VR).
• Improving agility & automation: Operators can update network policies in real-time to adapt to
user demands.
• Facilitating integration with NFV (Network Function Virtualization): Together, SDN + NFV
enable flexible deployment of network functions in software.

Key benefits in 5G
• Flexible & programmable networks Better resource utilization Simplified network management
Dynamic traffic engineering Supports edge computing & MEC (Multi-access Edge Computing).

SDN works in 5G architecture

• SDN Controller: Acts as the brain — communicates with underlying network devices (routers,
switches) via open protocols like OpenFlow.
• Northbound APIs: Allow applications to program the network behavior dynamically.
• Southbound APIs: Connect the controller to network devices.
• Integration with Core & RAN: SDN can be used both in the 5G Core Network (for routing, mobility
management, etc.) and in the Radio Access Network (RAN) for dynamic spectrum and resource
allocation.

Figure.2.13. SDN Architecture

Challenges
• Security: Centralized control is a potential attack target.
• Interoperability: Integrating SDN with legacy systems can be complex.
• Scalability: Large-scale 5G deployments need highly scalable SDN controllers.
• Standardization: Ensuring consistent standards across vendors.
Edge Computing
• Edge computing is about bringing the services closer to the location where they are
to be delivered. A service here includes computing power and memory needed for e.g.
running a requested application.
• Edge computing therefore aims to push applications, data and computing power
(services) away from centralized points (central data centers) to locations closer to
the user (such as distributed data centers). The goal is both to achieve a lower latency
and to reduce transmission costs.
• Applications that use high data volumes and/or require short response times, e.g. VR
gaming, realtime facial recognition, video surveillance etc. are some candidates that
could benefit from Edge computing.
• A lot of work in the industry around Edge computing has been done on the application
platform for edge applications and related APIs, e.g. by an ETSI Industry Specification
Group called MEC (Multi-access Edge Computing).
• In 3GPP however, the focus when it comes to edge computing has so far been
concerned with the access and connectivity aspects. This may change in the future
releases as new work is started 3GPP does not specify any special solutions or
architecture for Edge computing.
• Instead 3GPP defines several general tools that can be used to provide an efficient
User Plane path. These tools, most of which have already been described earlier in this
chapter, are not specific to Edge computing but they can be used as enablers in
deployments of Edge computing.
• The main tools for UP path management are listed below:
• UPF selection
• Selective traffic routing to DN
• Session and Service Continuity
• AF influence on traffic routing
• Network capability exposure
• LADN

Mobile Edge Computing

• Mobile Edge Computing (MEC) in 5G is a key concept that brings computing resources and
services closer to the end users, at the edge of the mobile network rather than relying entirely
on centralized cloud data centers. Let’s break it down in a clear way:
• MEC is an architecture that provides cloud computing capabilities within the Radio Access
Network (RAN), typically close to the base stations. Instead of sending all data to a faraway
cloud, some computation, storage, and processing happen near the user’s device.

MEC important for 5G

5G networks aim to deliver:
• Ultra-low latency (1–10 ms)
• High bandwidth
• Massive device connectivity
To achieve ultra-low latency and real-time responsiveness, it’s essential to process data closer to where
it’s generated — which is exactly what MEC does.

Key Benefits of MEC in 5G

• Reduced Latency: Data doesn’t need to travel far, improving response times.
• Improved Bandwidth Efficiency: Local processing reduces the load on the core network.
 • Context Awareness: MEC servers can use local information (like network conditions) to
optimize services.
• Enabling New Applications: Supports AR/VR, autonomous vehicles, industrial IoT, and smart
cities.

Example Use Cases

• Autonomous Vehicles: Real-time processing of traffic data at the roadside.
• Smart Factories: Instant monitoring and control of robots and machinery.
• Video Analytics: Real-time video surveillance without sending streams to a distant cloud.
• Gaming & AR/VR: Cloud gaming with low latency, or augmented reality with real-time object
detection.

MEC Integrates with 5G

• MEC nodes are deployed alongside 5G base stations.
• They interact with the 5G core network to provide local breakout of traffic.
• Virtualization technologies (like NFV and SDN) help deploy and manage MEC resources flexibly.

Challenges
• Deployment Costs: Need for new infrastructure at multiple edge sites.
• Security: More distributed architecture requires strong security measures.
• Standardization: Ongoing work by ETSI MEC, 3GPP, and other bodies.
• Interoperability: Integration with legacy systems and multiple vendors.

Key Standards & Bodies

• ETSI MEC ISG: Provides architecture and API standards.
• 3GPP: Defines how MEC interacts with the 5G core.
• Open-source initiatives: OpenStack, Kubernetes are often used to orchestrate edge nodes.

Quality of Service requirements

QoS in 5G
Quality of Service (QoS) in 5G refers to the ability of the network to provide different levels of
service quality to different types of traffic — based on their specific requirements for latency,
reliability, throughput, and priority.
This is crucial because 5G must simultaneously support diverse use cases, such as:
• Enhanced Mobile Broadband (eMBB)
• Ultra-Reliable Low Latency Communications (URLLC)
• Massive Machine-Type Communications (mMTC)

Key QoS Requirements

Here are the main QoS requirements in 5G:
1. Latency
• eMBB: 10–20 ms end-to-end latency.
• URLLC: as low as 1 ms end-to-end latency.
• Low latency is critical for applications like autonomous driving, remote surgery, or industrial
automation.
2. Reliability
• For URLLC: Up to 99.999% reliability.
 • This means extremely low packet loss, ensuring mission-critical data gets through.
3. Data Rate
• Peak data rates: up to 20 Gbps downlink, 10 Gbps uplink (under ideal conditions).
• User-experienced data rate: 100 Mbps or more for eMBB in dense urban areas.
4. Capacity & Connection Density
• mMTC: support up to 1 million devices per square kilometer.
• Needed for IoT devices, sensors, and smart city applications.
5. Availability
• High network availability is expected — especially for mission-critical apps like public safety or
industrial control.

5G ensure QoS
5G uses:
• QoS Flow: The basic unit for QoS in 5G — each flow has a QoS Identifier (QFI) with parameters
like priority, packet delay budget, and packet error rate.
• 5QI (5G QoS Identifier): Predefined or standardized QoS values for common services.
• Network Slicing: Logical network partitions with customized QoS for different use cases.
• Dynamic resource allocation: Via the 5G NR (New Radio) scheduler to meet real-time QoS
demands.
• Example mappings
Packet Packet
Use Case 5QI Priority Delay Error
Budget Rate
eMBB (video streaming) 9 Medium 300 ms 10⁻⁶
URLLC (autonomous driving) 85 High 10 ms 10⁻⁵
mMTC (IoT sensor) 75 Low 300 ms 10⁻³