CH A RTI N G TH E FUTU R E O F I N N OVATI O N | VO LU M E 102 I 2020–02

ERICSSON
TECHNOLOGY

CTO TECH TRENDS

CREATING INTELLIGENT
DIGITAL INFRASTRUCTURE

CRITICAL IOT
CONNECTIVITY
FOR INDUSTRY

INTEGRATED ACCESS
AND BACKHAUL
IN 5G NR NETWORKS
 CONTENTS ✱

Customer and partner interaction

BSS exposure layer

08 5G BSS: EVOLVING BSS TO FIT THE 5G ECONOMY

management
Intelligence
Managing complex IOT value chains and supporting new business
models requires more sophisticated business support systems (BSS)
Catalog
than those that communication service providers have used in the past. 08 Order capture and fulfillment

5G-evolved BSS enable smooth collaboration between connectivity

providers, service creators, partners, suppliers and others. Party
Charging Mediation Billing
Billing
management
20 OPTIMIZING UICC MODULES FOR IOT APPLICATIONS
= Decoupling and integration
The ability to deliver low-cost Internet of Things (IoT) devices on a mass scale
is at risk of being hampered by the high cost of the universal integrated circuit cards (UICC)
currently required to provide connectivity. Until a less costly alternative becomes available,
the IoT requires workarounds that either lower device cost or justify the price of UICCs
by leveraging more of their capabilities.
20

30 FEATURE ARTICLE
Future network trends: Creating intelligent
digital infrastructure
The vision of a fully digitalized, automated and programmable world of connected
humans, machines, things and places is well on its way to becoming a reality.
In his annual technology trends article, our CTO Erik Ekudden explains the seven
technology trends that are most relevant toslices
Network
HW capability
the network platform’s evolution
to become the platform for innovation to meet any societal or industrial need.
exposures
30
Zero-touch orchestration Business
intent

Gaming HW capability
AR/VRB APP APP control

E-MBB xNF xNF

40 THE FUTURE OF CLOUD COMPUTING: HIGHLY DISTRIBUTED
WITH HETEROGENEOUS HARDWARE
Cloud computing is being shaped by the Automotive
combination of the growing popularity
APP 40
xNF
of distributed solutions and increased reliance on heterogeneous hardware capabilities.xNF APP
APP APP APP

As the role of distributed computing in cloud computing continues to expand, network

SmartNICs

Manufacturing
operators, who have large, distributed systems already in place,
APP have a golden opportunity
xNF
xNF
to become major cloud players. Internet of xNF
Public clouds
Things PMEM APP xNF

52 CRITICAL IOT CONNECTIVITY – IDEAL FOR Distributed sites

xNF Central sites
TIME-CRITICAL INDUSTRIAL COMMUNICATIONS
Fixed access (edge/regional cloud) xNF: telco Virtual Network Function or
Access
Critical IoT connectivity is ideal for a wide range of Internet ofsites (edge cloud)
Things Cloud-native Network Function
APP: Third-party application
use cases across most industry verticals. Mobile network operators
are uniquely positioned to address the time-critical communication

52
needs of individual users, enterprises and public institutions by
leveraging their existing assets and new technologies in a
systematic fashion.

64 INTEGRATED ACCESS AND BACKHAUL

– A NEW TYPE OF WIRELESS BACKHAUL IN 5G
Integrated access and backhaul (IAB) is an advanced concept in 5G that shows significant
promise in addressing the challenge of wireless backhaul of street sites. IAB has several
advantages compared with other backhaul technologies, and if used properly, it could

64
become an essential backhaul solution for 5G NR networks.

#02 2020 ✱ ERICSSON TECHNOLOGY REVIEW 5

 ✱ EDITORIAL EDITORIAL ✱

Ericsson Technology Review brings you

insights into some of the key emerging
innovations that are shaping the
future of ICT. Our aim is to encourage
an open discussion about the potential,
practicalities, and benefits of a wide range
BRIDGING THE GAP
BETWEEN PHYSICAL
of technical developments, and provide
insight into what the future has to offer.

address
Ericsson
SE -164 83 Stockholm, Sweden
Phone: +46 8 719 00 00
AND DIGITAL REALITIES
publishing
All material and articles are published on the
Ericsson Technology Review website:
www.ericsson.com/ericsson-technology-review
■ the key role that connectivity plays in our daily
lives has never been more obvious – not only for GREATER AGILITY Another exciting 5G advancement that we present
in this issue is integrated access and backhaul (IAB),
publisher
Erik Ekudden
each of us as individuals but also for countless
enterprises around the globe. Thankfully, despite
AND SPEED WILL BE an innovative concept that shows significant promise
in addressing the challenge of wireless backhaul of

editor
the sudden, dramatic changes in our behavior in
early 2020, networks all around the world have
ESSENTIAL street sites.

Tanis Bestland (Nordic Morning) proven to be highly resilient. All seven of these trends serve as a cornerstone in We hope you enjoy this issue of our magazine
the development of a common Ericsson vision of and we’d be delighted if you share it with your
editorial board
Håkan Andersson, Magnus Buhrgard,
At Ericsson, we’re committed to ensuring that the what future networks will provide, and what sort of colleagues and business partners. You can find
Dan Fahrman, John Fornehed, Kjell Gustafsson, network platform continues to improve its ability technology evolution will be required to get there. both PDF and HTML versions of all the articles at:
Jonas Högberg, Johan Lundsjö, to meet the full range of societal needs as well as www.ericsson.com/ericsson-technology-review
Mats Norin, Håkan Olofsson, Patrik Roseen, supporting enterprises to stay competitive in the This issue of the magazine also includes five
Anders Rosengren, Robert Skog,
Gunnar Thrysin and Sara Kullman
long term. The ability to bridge distances and make additional articles highlighting some of our
it easier to efficiently meet needs in terms of resource latest research in the areas of cloud computing,
f e at u r e a r t i c l e utilization, collaboration, competence transfer, status the Internet of Things (IoT) and 5G advancements.
Future network trends: verification, privacy protection, security and safety The cloud computing article is particularly
Creating intelligent digital infrastructure is of utmost importance. Greater agility and speed noteworthy, as it explains how we think network
by Erik Ekudden
will be essential. operators can best manage the complexity of
art director future cloud deployments and overcome
Liselotte Stjernberg (Nordic Morning) My 2020 technology trends article, on page 30 technical challenges.
of this issue of the magazine, explains my view
project manager of the ongoing evolution of the network platform The first IoT article in this issue explains how critical
Susanna O’Grady (Nordic Morning) in terms of the key needs that are driving its IoT connectivity can be used to address time-critical
evolution and the emerging capabilities that needs in areas such as industrial control, mobility
l ay o u t
will meet both those and other needs. automation, remote control and real-time media,
Liselotte Stjernberg (Nordic Morning)
while the second one tackles the challenge that
i l lu s t r at i o n s The first three trends all relate to bridging the gap today’s universal integrated circuit cards (UICC)
Jenny Andersén (Nordic Morning) between physical reality and the digital realm – that is, present to IoT growth.
delivering sensory experiences and utilizing digital
subeditors representations to make the physical world fully With regard to 5G advancements, our BSS
Ian Nicholson (Nordic Morning)
Paul Eade (Nordic Morning) programmable. The emerging capabilities that I have article explores how 5G-evolved BSS can help ERIK EKUDDEN
highlighted this year are non-limiting connectivity, communication service providers transform SENIOR VICE PRESIDENT,
is s n : 0 014 - 0171 pervasive network compute fabric, trustworthy themselves from traditional network developers CHIEF TECHNOLOGY OFFICER AND
Volume: 102, 2020 infrastructure and cognitive networks. to service enablers and ultimately service creators. HEAD OF GROUP FUNCTION TECHNOLOGY

ERICSSON TECH NOLOGY REVIEW ✱ #02 2020 #02 2020 ✱ ERICSSON TECHNOLOGY REVIEW 7
✱ BSS IN THE 5G ECONOMY BSS IN THE 5G ECONOMY ✱

5G BSS:

Evolving BSS
A) Network developer B) Service enabler C) Service creator

Customer Customer Customer

to fit the IoT IoT Device

5G economy
CSP CSP provider CSP provider CSP manufacturer

SIM Device SIM Device

manufacturer manufacturer manufacturer manufacturer

Figure 1 The evolving role of the CSP in the IoT ecosystem

5G offers communication service providers an unprecedented opportunity
to enhance their position in the value chain and tap into new revenue into account, and mechanisms must be established Network developer, service enabler
streams in a variety of industry verticals. A successful transition will require to manage the relationships between them. or service creator?
Examples of new stakeholder groups that need Looking ahead, the capabilities that a CSP needs
business support systems (BSS) that are evolved to fit the 5G economy. to be considered in the 5G/IoT business context in its BSS solution will depend on the role it plays
include: – or aims to play – in the IoT ecosystem. Figure 1
illustrates the three role types: network developer,
❭ Enterprises and industry verticals that require service enabler and service creator.
solutions beyond telecoms In the traditional network developer role, a CSP
❭ New types of suppliers such as IoT device acts solely as a cellular connectivity provider by
The rapidly expanding Internet of Things ■ One of the primary roles of business support providers and suppliers of eSIM (embedded offering solutions such as radio, core network and
JAN FRIMAN,
(IoT) and all the new capabilities available systems (BSS) is to manage a CSP’s relationships SIM) and related technologies communication services. In this role, the CSP’s
MICHAEL NILSSON,
in 5G have opened up a wealth of opportunities with its stakeholders by keeping track of ❭ Platform providers that specialize in specific IoT business models are consumer focused. Its role in
ELISABETH MUELLER
for communication service providers (CSPs) agreements, handling orders, generating reports, or edge clusters or groups of use cases such as the IoT ecosystem is limited.
beyond their traditional markets, particularly sending invoices and so on. In the past, these massive and broadband IoT platforms, industrial In the service enabler role, the CSP extends its
in verticals such as automotive, health care, stakeholders were generally limited to consumers, IoT platforms and content data networks services by incorporating additional capabilities
agriculture, energy and manufacturing. resellers, partners and suppliers. In the 5G/IoT ❭ Integrators that specialize in specific verticals such as cloud/edge and IoT enablement and shifts
To monetize them, CSPs will need to meet business context, though, more complex such as asset management, mission-critical focus to business customers and industry verticals.
the expectations of a broader range ecosystems are arising that BSS must evolve to services or automotive that combine The CSP becomes a service enabler for 5G and the
of stakeholders and be able to handle support. To do so, the requirements of a larger, capabilities from multiple stakeholders to IoT, acting as a supplier of connectivity and platform
complex ecosystems. more diverse group of stakeholders must be taken address consumer needs. services. As a service enabler, the CSP’s business

8
2 E R I C S SEO
RNI CTSESCOHNNTOELCOHGNYORL E
OVGI Y
EWR E✱V I EMW
A R✱
C H#2062, 2 0 2 0 # 2 0H2206 ,✱
M 0A2R C 0 I C✱S SEO
2 0E2R RNI CTSESCOHNNTOELCOHGNYORL E
OVGI Y
EWREVIEW 9
3
✱ BSS IN THE 5G ECONOMY BSS IN THE 5G ECONOMY ✱

models are extended to business-customer focused – that is, enabling the BSS to handle traffic and BSS solution level Capabilities
with respect to 5G IoT. a large number of devices at IoT scale.
In the service creator role, the CSP transitions In terms of functionality, the BSS enhancements 5G enabled • 5G service-based interface (SBI) support (charging function)
• Network slicing support in BSS and OSS
from being a connectivity and platform provider to required by service enablers include: • Classic roaming partners
creating new digital services and collaborating • Containerization and microservices
beyond telecoms to establish digital value systems. ❭ Automation of full life-cycle management for
• Common technology stack
As a service creator, the CSP partners with suppliers devices/IoT resources supported by flexible
to deliver new services all the way up to full IoT orchestration, including exposure of services I OT and edge • ID management and correlation
for managing relationships with business monetization • Life-cycle management for IoT devices
solutions, taking on the roles of integrator,
• Business customer and 5G/IoT enterprise management
distributor or co-seller. customers
• Charging in multilevel hierarchies
❭ Support for batch orchestration, flexible supply • Supply agreements
BSS for all three CSP roles agreements and contracts for non-telco • Flexible orchestration of ordering processes
Traditional BSS support the CSP in the network services with associated charging models • Service exposure for device management
developer role, in which the CSP charges for voice, ❭ Service exposure of network capabilities, so • Open API exposure
• Continuous integration/continuous delivery (CI/CD) for service exposure
text and data services based on consumption or that IoT providers can bundle their offerings • Enterprise self-care
subscription level. The main requirements for with connectivity and sell them on to their • Multiparty charging and best-effort charging
these BSS are: customers • Private networks
❭ Service exposure of BSS and OSS capabilities • Platform partnerships
❭ Customer management, traditional partner • Contract for non-telco services (IoT/edge enabled)
to enable efficient ordering processes,
business (roaming partners), charging and • Charging models for non-telco services
especially with regard to the management of • Multi-tenancy
billing, and finance modules mass subscriptions. • Charging and billing on behalf of
❭ Order capture and order execution for new • Location-aware services
telco subscriptions and/or add-on offerings Supporting a CSP in the service creator role, where • Blockchain for smart contracting
❭ Charging and balance/quota management the full life cycle of partners must be taken into • Service Level Agreement (SLA) management
in BSS, as well as mediation account, requires BSS with further functional
Full 5G ecosystem • Partner relationship management
❭ Interaction with operations support systems extensions. The stakeholder ecosystem of service • Partner catalog
(OSS) for network provisioning. creators is significantly more complex, as the • Partner revenue sharing
customer base broadens to include verticals and the • Reconciliation and settlement
Evolving BSS to support a CSP in a service enabler CSP starts offering full solutions beyond telecoms. • Flexible billing
role requires a shift in focus to the needs of As a result, BSS for service creators must include • Platform as a service and distributed cloud
• Edge platform services
enterprise customers and IoT use cases. The BSS extensive and flexible partner relationship
• Multi-access edge computing (MEC)
must be transformed into a system that is able to management. Supply chain management is • BSS as a service
monetize IoT/5G platforms and edge deployments, especially important. • Continuous monitoring
which requires significant changes in both the The capacity to expose network capability as well • Artificial intelligence and machine-learning automation
functional and non-functional space. In the non- as BSS and OSS capabilities is critically important to • CI/CD
functional space, this mainly involves scalability a CSP’s ability to deliver on service creation beyond
Table 1 Key capabilities of the three BSS solution levels

telecoms, so that partners can develop tailored sharing all require extended billing and
applications and deploy them on the operator’s reconciliation functionality.
Terms and abbreviations infrastructure.
API – Application Programming Interface | BSS – Business Support Systems | CSP – Communication Finally, the new business models available to BSS solution levels and key capabilities
Service Provider | IoT – Internet of Things | ODA – Open Digital Architecture | OSS – Operations Support CSPs as service creators require new monetization Table 1 organizes and sequences key BSS
Systems | SBI – Service-Based Interface | SDK – Software Development Kit | SLA – Service Level Agreement models for charging and billing. For example, capabilities based on technical dependencies and/or
multiparty charging, revenue sharing and profit level of complexity. One by one, these capabilities

10
4 E R I C S SEO
RNI CTSESCOHNNTOELCOHGNYORL E
OVGI Y
EWR E✱V I EMW
A R✱
C H#2062, 2 0 2 0 # 2 0H2206 ,✱
M 0A2R C 0 I C✱S SEO
2 0E2R RNI CTSESCOHNNTOELCOHGNYORL E
OVGI Y
EWREVIEW 11
5
✱ BSS IN THE 5G ECONOMY BSS IN THE 5G ECONOMY ✱

BSS, with similar interfaces to surrounding systems.

SCALABILITY ALONE IS NOT The BSS architecture in Figure 2 is presented in the
ENOUGH TO HANDLE MASSIVE Open Digital Architecture format [1]. It is divided
into party management, core commerce
AMOUNTS OF DEVICES management, intelligence management, production Social Customers Business Developers Apps
and engagement management. Production includes media customers
add on to each other, continuously increasing BSS the southbound application programming interface
maturity and transforming the BSS into a system (API) layer to the network infrastructure, IoT
capable of supporting all the new use cases and platforms, cloud/edge and OSS, while engagement Northbound API
business models that characterize the 5G/IoT management includes the northbound API layer to
ecosystem. customers and partners.
The first evolution step – ‘5G enabled’ in Table 1 5G and the IoT place several challenging Party management

management
– provides support for new 5G standards and requirements on new capabilities in the BSS

Engagement

management
Intelligence
concepts, which enables a drastic increase in data architecture that are not directly visible at a high
transmission throughput while maintaining focus on level. All functional areas are affected by the 5G
traditional consumers. Applying containerization evolution and are extended to support the new Core commerce management
and a common technology stack will assure the requirements and possibilities, most notably in the
scalability of the BSS solution to meet the increased areas of mass-device management, device and
throughput demands of the network. resource life-cycle management, subscription
At the next solution level, IoT and edge monetization, management, charging models for non-telco Production
the focus shifts to business customers. These new services and multiparty charging.
capabilities enable the CSP to provide extended
support for enterprises when it comes to 5G and IoT IoT-scale mass-device management
use cases by covering IoT device management, The sheer number of connected devices in the 5G/ Southbound API
support for non-telco service charging and multi- IoT world is a major challenge for BSS to manage.
party charging as well as IoT and/or edge-platform While current BSS architectures are scalable, they
= Decoupling and integration
monetization. In addition, service exposure enables will be too costly for IoT use cases due to the large
self-service for enterprises along with application data footprint and processing need of each device.
development for the optimization of IoT devices. Scalability alone is not enough to handle massive
The number of 5G/IoT use cases that the CSP is able amounts of devices. To address this, 5G-evolved
to support increases drastically at this stage. BSS must have a persistence and management
The addition of partner capabilities at the full model that is lightweight enough to allow a large Comm. EPC/ Cloud/
5G ecosystem level allows the CSP to address totally number of devices to use the same footprint as one Mediation Policies IoT OSS
services 5G Core edge
new customer segments beyond telecoms and traditional device. This can be addressed using
provide industry-specific solutions to verticals. concepts such as herding, where each individual
A CSP can create new services (even deliver BSS device only requires a minimal data footprint. Figure 2 5G reference architecture for BSS
as a service), and offer these services on a marketplace The behavior of each individual device is
to reach new segments of business customers. determined by the herd configuration, which is
The multitude of partnerships require support for a single specification per herd. monitor the state of the device throughout its The calculation of charges related to IoT devices
new business models that allow flexible charging, life cycle is not sufficient. For example, contracts is also complicated by the fact that the state of the
revenue sharing and billing. Life-cycle management of that cover large herds of devices are likely to be device can influence the charged party. One example
IoT devices and resources based on recurring charges per active device. of this challenge is IoT devices that are mounted in
5G reference architecture for BSS Managing the life cycles of IoT devices and In these scenarios, the aggregated numbers of vehicles at a factory. The factory personnel will likely
From a high-level architectural viewpoint, BSS in resources is another significant challenge for BSS. In devices per state become key parameters in the want to test that the device is working before
the 5G/IoT ecosystem closely resemble traditional many emerging IoT applications, the ability to calculation of charges. shipping the vehicle to the reseller. The reseller may

12
6 E R I C S SEO
RNI CTSESCOHNNTOELCOHGNYORL E
OVGI Y
EWR E✱V I EMW
A R✱
C H#2062, 2 0 2 0 # 2 0H2206 ,✱
M 0A2R C 0 I C✱S SEO
2 0E2R RNI CTSESCOHNNTOELCOHGNYORL E
OVGI Y
EWREVIEW 13
7
✱ BSS IN THE 5G ECONOMY BSS IN THE 5G ECONOMY ✱

ONE EFFECTIVE BSS with a unique opportunity to determine which

charging, balance management and aggregation
APPROACH IS TO EXPOSE functions must be performed, and use this Customer and partner interaction
APIs AND TOOLS THAT ALLOW knowledge to monetize the usage of the 5G network.
For instance, the BSS can monitor allowances and
PARTNERS ... TO ONBOARD balances in real time, if so required by a partner BSS exposure layer
agreement, or decide to postpone the rating and
AND MANAGE DEVICES

management
Intelligence
balance management to a near real-time
then want to demonstrate the service the device asynchronous flow.
Catalog Order capture and fulfillment
provides to prospective buyers, before a consumer Allowing the BSS to decide the importance and
ultimately buys the vehicle and starts using the risk level of each event based on agreements, Service
service. At each of these stages, the charged party Level Agreements (SLAs) and operator business Party
and charging model may be different depending on rules makes it possible to accommodate multiple Charging Mediation Billing
Billing
management
the state of the device. Overcoming such challenges charging models simultaneously. Among other
requires a BSS architecture that can provide up-to- things, this approach enables real-time monitoring
date state information per individual device or of individual device herds, while at the same time = Decoupling and integration
resource as well as aggregated information to the providing partner ratings for one or multiple
rating, charging and billing functions. involved partners in a continuous, near real-time,
flow for individual device sessions. Figure 3 Ericsson’s digital BSS implementation architecture
Subscription management for IoT devices
Subscription management is another area that must Charging models for non-telco services
evolve to fit the new 5G/IoT business context. 5G-evolved BSS must also support the management Multiparty charging access specific services, the event is followed up by a
Traditional BSS are built to manage consumer and monetization of services that are not traditional While traditional BSS are able to handle roaming post-session process to calculate and distribute the
subscriptions. They are not capable of handling the telco services, such as those for the IoT platform or partners and wholesale agreements, they are not charges/revenue share for the involved partners.
massive number of devices in IoT use cases in a cost- application hosting at the edge. In the past, BSS equipped to handle the dramatic increase in As a result, the relevant partners have access to
efficient manner. Subscription management in have traditionally relied on a well-defined set of different types of partner agreements in the 5G/IoT up-to-date information within seconds, rather than
5G-evolved BSS requires a high level of automation parameters provided through standardized ecosystem. The ability to handle a wide variety of at the end of the day or at the bill run as they would
and solutions that reduce the processing footprint to protocols, but this approach will not be sufficient partner agreements and support the onboarding of in traditional BSS.
onboard and manage devices, services and products. when entering the non-telco service arena. partners and related charging models will be crucial In 5G-evolved BSS, different events for the same
One effective approach is to expose APIs and tools To monetize on non-telco services, the 5G-evolved to CSPs’ ability to monetize on expected IoT growth service can have different charge or revenue share
that allow partners or even consumers to onboard BSS must have the flexibility to use previously and avoid becoming bit-pipe wholesalers. distribution. One-time fees, recurring charges or
and manage devices. unknown identifiers and parameters, especially In the 5G/IoT ecosystem, a single event that BSS usage fees can all have different distribution rules
To gain efficiency and minimize management, in the charging and billing systems. receive from the 5G core network can trigger a and include one or more partners. For example, it is
pools of services and products can be linked to herds The usage of a non-telco service can be monetized complex value chain that requires multiple parties to possible for an operator to charge a one-time fee
of devices, instead of applying individual services to using something as simple as a network slice be charged or share revenue. A CSP cannot rely on to a consumer and keep all of the revenue, while also
device relationships, which is the common practice identifier to determine how to aggregate and charge traditional techniques to handle this complexity – charging a recurring fee to the same consumer and
in BSS today. The service instances linked to herds for a service. In other instances, a much more doing so would mean postponing charging or splitting that revenue with a partner that provides
are kept to a minimal footprint and the majority of complex model must be used, involving multiple revenue share distribution until the bill run. the consumer device on a rental basis.
the parameters needed for processing can be kept on input parameters for each event to determine which To deliver up-to-date information to the relevant
specification level. This change will enable more party or parties should be charged and which partners, the CSP needs BSS that can process the Digital BSS architecture for 5G and the IoT
efficient processing in BSS and reduce the number charging model should be applied. Consequently, entire value chain as soon as any activity occurs that Figure 3 shows the key components of Ericsson’s
of scenarios that require mass provisioning. the charging and billing solution in 5G-evolved impacts them. This does not mean that everything digital BSS architecture. The color scheme indicates
Unlike traditional BSS, 5G-evolved BSS must be BSS must provide the flexibility to map and evaluate must be processed in real time, but rather that events the relationship between the components in this
able to capture and create the network charging data non-telco identifiers and other parameters at must be handled in an online asynchronous process. architecture and the functional ODA architecture
records (charging function). This task provides the configuration time. For example, when BSS grant consumers the right to shown in Figure 2.

14
8 E R I C S SEO
RNI CTSESCOHNNTOELCOHGNYORL E
OVGI Y
EWR E✱V I EMW
A R✱
C H#2062, 2 0 2 0 # 2 0H2206 ,✱
M 0A2R C 0 I C✱S SEO
2 0E2R RNI CTSESCOHNNTOELCOHGNYORL E
OVGI Y
EWREVIEW 15
9
✱ BSS IN THE 5G ECONOMY BSS IN THE 5G ECONOMY ✱

The front-end channels in the customer and partner Further down in the stack, the architecture is based architecture. Containerization, microservices and a 5G-evolved BSS enable smooth collaboration
interaction layer and the BSS exposure layer are on mini services, primarily to optimize footprint, common technology stack are common to all blocks. between connectivity providers, service creators,
deployed as a microservice architecture to facilitate performance and latency. partners, suppliers and others that results in the
business agility, scaling and the introduction of Table 2 maps out the 5G evolution areas in BSS efficient creation of attractive and cost-effective
customized solutions as per operator needs. to the main functional blocks in our digital BSS CSPs NEED BSS THAT services. Optimized information models and a high
ARE EVOLVED TO MANAGE degree of automation are required to handle huge
numbers of devices through open interfaces.
BSS functional block 5G evolution areas COMPLEX VALUE CHAINS Deployment in a cloud-native architecture ensures
Customer and partner • Catalog driven, omnichannel flexibility and scalability. It is important to keep the
interaction • B2C and B2B digital front end: customer/partner journeys Conclusion business logic, interfaces and information models
• B2C and B2B CPQ (configure, price and quote), frame contracts The 5G network evolution presents communication of 5G-evolved BSS flexible, so they can be adjusted
• B2B2X marketplace
service providers with the opportunity to transform to suit the value chains and business models of the
BSS exposure layer • OpenAPI exposure themselves from traditional network developers to different industry verticals.
• Loosely coupled principle service enablers for 5G and the Internet of Things, At Ericsson, we will continue to evolve our BSS
• SDK to support API aggregation and ultimately to service creators with the ability to offering to support our customers on their journeys
collaborate beyond telecoms and establish lucrative from network developers to service enablers, from
Catalog • Exposure for partner product creation
• Enhanced bundling with partner products
digital value systems. Along the way, this journey service enablers to service creators and beyond.
• Product models for network resources opens up substantial new revenue streams in As part of this work, we are also firmly committed
• Product models for enterprise products verticals such as industrial automation, security, to driving and contributing to relevant standards
• Partner catalog health care and automotive. To successfully in the BSS area and participating in open source
• Multi-device offerings capitalize on this opportunity, CSPs need BSS and developer communities to promote openness
that are evolved to manage complex value chains and interoperability.
Order capture • Ecosystem orchestration
and fulfillment • New business model support and support new business models.

Charging • Support for new charging trigger points

• Manage communication services at IoT scale
• Charging life-cycle management as a part of mass IoT device
and mass subscription life-cycle management
• Multiparty charging References
• Charging in hierarchies 1. TMA, Open Digital Architecture Project, available at: https://www.tmforum.org/collaboration/open-digital-
• Charging on behalf of architecture-oda-project/
• Non-telco service charge

Mediation • Call detail record generation for 5G

• Online mediation SBI -> diameter
Further reading
Party management • Extended B2B (supply agreements, non-telco contracts)
❭ Ericsson Technology Review, BSS and artificial intelligence – time to go native, January 2019, available at:
• Digital partner management
https://www.ericsson.com/en/reports-and-papers/ericsson-technology-review/articles/bss-and-artificial-
Intelligence management • SLA management intelligence-time-to-go-native
• Data lake ❭ Ericsson blog, Impacts of monetizing 5G and IoT on Digital BSS, October 29, 2019, Michael Fireman,
available at: https://www.ericsson.com/en/blog/2019/10/impacts-of-monetizing-5g-and-iot-on-digital-bss
Billing • Life-cycle management as a part of mass IoT device and
mass subscription life-cycle management ❭ Ericsson blog, Monetize 5G and IoT business models, October 7, 2019, Michael Fireman, available at:
• Multiparty billing https://www.ericsson.com/en/blog/2019/10/monetize-5g-and-iot-business-models
• Billing on behalf of
• Revenue sharing ❭ Ericsson, Telecom BSS, available at: https://www.ericsson.com/en/portfolio/digital-services/digital-bss
• IoT partner settlements ❭ Ericsson, Digital BSS, available at: https://www.ericsson.com/en/digital-services/offerings/digital-bss

Table 2 Prioritized 5G evolution areas in the main BSS functional blocks

16
10 E R I C S SEO
RNI CTSESCOHNNTOELCOHGNYORL E
OVGI Y
EWR E✱V I EMW
A R✱
C H#2062, 2 0 2 0 # 2 0H2206 ,✱
M 0A2R C 0 I C✱S SEO
2 0E2R RNI CTSESCOHNNTOELCOHGNYORL E
OVGI Y
EWREVIEW 17
11
✱ BSS IN THE 5G ECONOMY

within Business Area Digital She joined Ericsson in 2006

Services. Nilsson joined when LHS in Frankfurt was
the authOrs

Ericsson in 1990 and has acquired to complement the

extensive experience from Ericsson BSS offerings with a
the telecommunications billing system. Since then she
area in support and has taken on many different
verification, radio, core and roles within the company,
transmission network design including system design,
system management and
Jan Friman solution architecture in all
◆ is an OSS/BSS expert BSS areas. Mueller holds an
in the architecture and M.Sc. in mathematics from
technology team within Johannes Gutenberg
Business Area Digital University in Mainz,
Services. Since joining Germany, along with several
Ericsson in 1997, he has held patents in the BSS area.
various OSS/BSS-related
positions within the
company’s R&D, system and BSS product
management and strategic development. Since 2012, he
product management has held the position of chief
organizations. Friman holds architect for next generation
an M.Sc. in computer science BSS development.
from Linköping University,
Sweden. Elisabeth Mueller
◆ is an expert in BSS
Michael Nilsson end-to-end systems whose
◆ is a BSS expert in the current work focuses on
solution architecture team 5G/IoT BSS architecture.

18
12 E R I C S SEO
RNI CTSESCOHNNTOELCOHGNYORL E
OVGI Y
EWR E✱V I EMW
A R✱
C H#2062, 2 0 2 0
✱ UICC MODULES AND THE IoT UICC MODULES AND THE IoT ✱

OPTIMIZING

UICC modules
While the advanced features of UICC modules Optimization
continue to provide considerable value in mobile A typical operator profile on a 3GPP consumer
phone applications, most of them are superfluous mobile phone is up to tens of kilobytes; the average
in IoT applications. In light of this, the industry IoT sensor only requires 200-300 bytes. And of all
is working to find a less sophisticated solution the functionality that a UICC can provide, an IoT
that is more appropriate for applications that device only really needs the Universal Subscriber

for IoT
require massive numbers of devices in price- Identity Module application and the remote SIM
sensitive environments. Industry alignment on provisioning (RSP) application, which allows
such a solution is expected to be a challenging and remote provisioning of subscriber credentials
time-consuming process, however, due to the fact (also known as operator profiles).
that the IoT area is fragmented into many different One good way to significantly reduce the footprint

applications
verticals, application areas and use cases. of the UICC is to optimize the operator profile and
Ericsson is fully committed to supporting the the necessary software environment within the
long-term, industry-aligned solution. In the meantime, UICC module. Doing so not only saves storage in the
however, it is vital to find workarounds to ensure device but also reduces energy consumption during
that the cost of UICCs does not stifle IoT growth. over-the-air download. Further size reduction
While the definitive solution to the question of of the device may be achieved when the UICC is
what should replace the UICC is hard to predict, completely integrated into the baseband modem
two mid-term workarounds are clear: the complexity or application processor (integrated UICC or
of using UICCs and leveraging their excess capacity iUICC [2]). This simplified and integrated solution
to generate additional value. could work effectively for use cases that require
The UICCs used in all cellular devices today are complex and powerful low-cost, simple, secure and low-power IoT devices
Reducing the complexity of using UICCs in high volumes.
minicomputers capable of much more than most Internet of Things (IoT) There are three main approaches to reducing the The use of an iUICC requires an effective
applications require. Until a simpler and less costly alternative becomes complexity of using UICCs in IoT applications: RSP protocol [3, 4] that makes it possible to
optimization, usage of 3GPP standardized change subscription credentials. Current RSP
available, it makes sense to find ways to reduce the complexity of using certificate-based authentication, and standards are too complex for iUICCs for many
them and use their excess capacity for additional value generation. virtualization. reasons, including their use of HTTPS

BENEDEK KOVÁCS, UICCs are used today to facilitate network dozens of Java-based applications in parallel and Definition of key terms
ZSOLT VA JTA , connection in all 3GPP user equipment – support powerful remote management operations. Identity describes the link between the identifier of an entity and the credentials that it uses to prove
ZSIGMOND PAP mobile phones, IoT devices and so on. Backward-compatibility is provided by running that it is the rightful owner of the identity.
a network service application on UICC modules,
■ The most important tasks of UICC modules – which can emulate the file system for storing First used in Finland in 1991, the original subscriber identity module (SIM) was a smart card with
commonly referred to as SIM cards – in today’s necessary credentials and old-school smartcard a protected file system that stored cellular network parameters. It was designed to connect
mobile networks are to store network credentials protocols, extended with features such as enhanced expensive user equipment – mobile phones – with expensive subscriptions to the cellular network.
and to run network security and access security, extended telephone register and operator
applications in a secure and trusted environment. logo image. The interface between the UICC module When it became clear that smart cards did not have the capacity to provide an adequate level of security
In addition, they are also capable of storing a large and the user equipment (devices) is standardized, in next-generation cellular networks, they were replaced with universal integrated circuit cards (UICCs)
amount of extra information and running multiple which enables operators to run value-added – minicomputers equipped with general microprocessors, memory and strong cryptographic
co-processors [1].
toolkit applications. A UICC’s own operating applications, such as mobile wallet or mobile lottery,
system provides a full Java environment. It can run on the UICC module.

20
2 ER
ERICS S IOCNS S
TEOCNHTNEO
CLHONGOYL O
RGE VYI ERW
E V✱
IEW
A P✱
R I L# 1042, 2 0 2 0 #P
A 2 I2L01240, 2 ✱
0R 0 2 0E R✱
ICS
ERS IOCNS S
TOE CNHTNEO
CLHONGOYL O
RGE VYI ERW
EVIEW 21
3
✱ UICC MODULES AND THE IoT UICC MODULES AND THE IoT ✱

authentication has been performed. According to when an IoT device needs to manage multiple
Device owner/user the 3GPP, authentication in private networks operator profiles – a circumstance that will become
such as Industry 4.0 solutions may rely entirely increasingly common, according to an analysis
on certificate-based solutions such as Extensible carried out by the GSMA [5].
Authentication Protocol over Transport Layer The disadvantages of virtualization are similar to
Security. Without a UICC for securely storing those of certification-based solutions. Most notably,
LwM2M-based and operating on secret long-term credentials certification is harder when a trusted environment
Internet
secure communication for network access authentication, another is integrated with the rest of the device compared
secure environment with secure storage with using an isolated UICC or eUICC.
IoT Provisioning Mobile solution is needed.
platform
HTTPS
server
(SM-DP+)
network
operator
For certain applications a lower level of security
might be accepted. The value of the data that the
OPERATORS MUST
LPAdv SIM alliance profile LPApr
IoT device provides or handles, in relation to the DECIDE WHICH SOLUTION
cost of the IoT device, determines the required
security level of the secure environment for protecting
IS MOST SUITABLE FOR
LPA split network access authentication credentials. In the ANY GIVEN APPLICATION
IoT device with
case of a UICC being used, it determines the
cellular module
realization of the UICC functionality. For some Generating additional value from the UICC
low-cost constrained IoT devices, a realization Experience shows that it is significantly less
Figure 1 Remote provisioning using IoT-optimized technology using a hardware-isolation-based trusted execution expensive to limit a protected and certified
environment may be acceptable. As there is no manufacturing environment to a dedicated
universal and perfect solution, operators must hardware module such as a UICC than to ensure
(Hypertext Transfer Protocol Secure) and reliance or connectivity management server. The device decide which solution is most suitable for any given that all the software running in the mobile
on SMS support. HTTPS is typically not part of the management protocol stack (Open Mobile Alliance application. It is likely that the UICCs and eUICC- equipment can be trusted. In light of this, we believe
protocol stack of constrained low-power IoT devices. (OMA) LwM2M [1], for example) handles the based solutions will remain the technology of choice that communication service providers will continue
Instead, these devices use a stack with Constrained communication between the two LPA parts. in public networks for the next few years. using UICC modules for at least the next 5-10 years.
Application Protocol (CoAP), Datagram Transport Profile protection is still end-to-end between During this period, it makes sense to exploit the
Layer Security (DTLS) and User Datagram the iUICC/embedded-UICC (eUICC) and the Virtualization potential of the UICCs to better support IoT
Protocol. In some cases, the Lightweight Machine- provisioning server (Subscription Manager-Data Virtualizing the UICC is yet another alternative applications by creating value-added services
to-Machine (LwM2M) protocol is used on top of Preparation – SM-DP+). that addresses the cost issue associated with for operators and enterprises. Three examples of
CoAP for device and application data management. UICC technology. One way to do this is to run this are using the UICC as cryptographic storage,
The use of only one stack keeps the cost of the Usage of 3GPP standardized a UICC environment in a virtual machine using it to run higher-layer protocol stacks,
device down. certificate-based authentication (or at least on a separated processor core) inside and using it as a supervisory entity.
Ericsson proposes utilizing the same protocol Another way to reduce the need for a UICC the application processor or the baseband modem.
stack for profile download and profile management is to use a network authentication mechanism Another approach is to store the operator profiles Using the UICC as cryptographic storage
as is used for device and application data different to the classical 3GPP Authentication in the security zone of the application processor, UICC modules were designed to serve as
management. Figure 1 illustrates how to achieve and Key Agreement (AKA). The use of certificates then download them to empty physical UICC cryptographic storage and are used today mainly
this by adapting the GSMA embedded-SIM is a classic solution used in the internet that may hardware on demand. for the storage of security credentials for 3GPP
solution for consumer devices for use with IoT easily fit into the existing network architecture The biggest advantages of these virtualization connectivity. We propose, in accordance with
devices. In this solution, the local profile assistant of an enterprise/service provider. In public solutions is flexibility and better utilization of GSMA IoT SAFE [1], that the UICC itself should
(LPA) is split into two parts. To reduce device 5G networks, authenticating with certificates existing hardware resources, while at the same time also be used as a crypto-safe for the IoT platform,
footprint, the main part of the LPA (including the is possible as a secondary authentication for a maintaining many of the advantages of current providing support to establish encrypted
use of HTTPS) is moved from the device to a device service using AKA, but only after primary network technology. These methods are particularly effective connection of the applications.

22
4 ER
ERICS S IOCNS S
TEOCNHTNEO
CLHONGOYL O
RGE VYI ERW
E V✱
IEW
A P✱
R I L# 1042, 2 0 2 0 #P
A 2 I2L01240, 2 ✱
0R 0 2 0E R✱
ICS
ERS IOCNS S
TOE CNHTNEO
CLHONGOYL O
RGE VYI ERW
EVIEW 23
5
✱ UICC MODULES AND THE IoT UICC MODULES AND THE IoT ✱

A generic IoT device has multiple identities for use The recently released GSMA IoT SAFE [8] offers
in multiple security domains. Every identity has at a solution where the UICC is utilized as a root of IoT device
least one identifier and credential, all of which must trust for IoT security. Here, an applet on the UICC/
be stored somewhere. Although there are multiple eUICC provides cryptographic support and storage Operator Sensor data
options, a hardware element that is powerful enough of credentials for establishing secure communication profile
to play the role of the root of trust is definitely needed. (for example, using DTLS) to an IoT service. The
PSK
The UICC is perfect for this role, as it is already used existing UICC management system (UICC OTA
as an identity for 3GPP networks, storing International mechanism) is used by the operator to establish UICC
Device and
Mobile Subscriber Identity, intensified charge- trusted credentials between the device and the IoT data

SIM toolkit
coupled device, Wi-Fi and OMA LwM2M [6] service. The GSMA IoT SAFE defines an application PSK management
Application (LwM2M
credentials along with dozens of other identifiers. programming interface for interoperability between LwM2M server)
The necessary trusted and certified environment SIM applets from different operators. client
and infrastructure are already available to manufacture Dev. ID BIP
the module, download and update its content and EFFECTIVENESS
carry out remote management as well.
To cover every aspect, UICC-based solutions COMES FROM ITS POWER- NIDD/SMS/USSD

require cooperation between the UICC ecosystem FUL CRYPTOGRAPHIC IMEI

NIDD/SMS SCEF
and the IoT device security subsystem (ARM Trust Radio modem /USSD
Zone [7], for example). ID and credential management CO-PROCESSORS
itself is device-independent, which saves development
cost and increases the security level. Additional Using the UICC to run
advantages of using UICC as a root of trust are: higher-layer protocol stacks Figure 2 IoT device with LwM2M client running on the UICC module, using NIDD
In addition to providing security and encryption
❭ it has its own local processor functions, UICC modules could also serve as
❭ it is usually equipped with powerful main application processors. Today, a low-cost, Since modem firmware is a closed environment, IP layer because it is running in the signaling
cryptographic co-processors sensor-like IoT device usually has at least three it is difficult to upgrade and to customize its protocol network of the operator. An additional advantage
❭ it comes with a powerful, standardized remote processors on board: one is on the UICC module, stacks (extending them with proprietary added of this approach is that it increases interoperability.
management subsystem (RMS) another runs inside the baseband modem, and a values). In addition, a small security hole in the There is a standardized way of upgrading the
❭ it is handled through a separate logistics chain. third – the application processor itself (sometimes protocol stack can be enough for a hacker to take communication stack in the UICC – it is even
combined) – collects data and hosts higher level control of the whole modem. possible to insert the communication stack into
The UICC can generate key-pairs and store communication stacks such as LwM2M, CoAP Alternatively, these higher-layer protocol stacks the operator profile. This does not completely
private keys for multiple security domains or MQ Telemetry Transport. can be moved to the UICC. Figure 2 depicts a block solve compatibility and interfacing problems,
effectively and securely. Effectiveness comes Shifting the higher-level communication stack diagram of a device, where the OMA LwM2M but a certified operator can handle these issues
from its powerful cryptographic co-processors, from the application processor to the UICC client runs on the UICC module and uses a non-IP on a higher security level to provide wider
while security is provided by the combination of module can lead to cheaper hardware and lower data delivery (NIDD) protocol connection to send solution matching.
the standardized RMS and the UICC’s ability to development costs, as well as providing a unique information to the device management system. In the simplest IoT devices, it might even be
run cryptography processes inside the module. approach to interoperability. As a result, some Running higher-level protocols in the UICC possible to run the actual IoT application on the
This means that the keys never leave the hardware modem manufacturers have implemented these module can improve security in several ways. UICC module. This would open for edge-computing
and therefore they cannot be exposed to the protocols inside the modem, running a complete For example, it is possible to run the LwM2M solutions in which simple tasks are executed on the
application. Not only does this architecture OMA LwM2M protocol stack in the baseband chip, stack over a NIDD connection [9] and even to allow device – data filtering to reduce the amount of data
provide security, it can also securely tie for example. While this may free up an external this code to execute on the UICC module instead being sent over the air, for example. Security can also
the 3GPP connectivity credentials and other application processor and speed up device of on the device processor. In this scenario, be improved if the binary is stored on the UICC
IoT certificates to each other. development, this solution is rather inflexible. command/control is never exposed on the instead of on the device application processor.

24
6 ER
ERICS S IOCNS S
TEOCNHTNEO
CLHONGOYL O
RGE VYI ERW
E V✱
IEW
A P✱
R I L# 1042, 2 0 2 0 #P
A 2 I2L01240, 2 ✱
0R 0 2 0E R✱
ICS
ERS IOCNS S
TOE CNHTNEO
CLHONGOYL O
RGE VYI ERW
EVIEW 25
7
✱ UICC MODULES AND THE IoT UICC MODULES AND THE IoT ✱

Using the UICC as a supervisory entity management system, data management system,
Zero-touch provisioning (ZTP) is yet another and so on). Several standardized technologies exist ZTP service
possibility for better utilization of the UICC to support this process but, unfortunately,
module. ZTP refers to the possibility of adding they are not connected into a working, efficient,
an identity to a device when required, with fully automated and cooperative system.
automatic setup of the working environment The most straightforward way to connect Mobile
Data Device Enterprise
(requiring manual intervention). different subsystems in a flexible and programmable Device vendor
management management
network UICC vendor
CRM operator
An effective automatic provisioning system way is to run a centralized service above or at the
requires remote provisioning management, same level as these subsystems. This ZTP service
key and credential storage, identity mapping of is connected to the 3GPP network (for instance
UICC modules and applications as well as strong to subscriber data management), to the SM-DP+
flexibility in case of operator profiles, but all of this system (usually operated by the UICC module IoT device
is far from enough. Provisioning of IoT devices is a vendor or an independent bootstrap operator),
complex, slow and costly procedure. Although there to the device management system and to the data Application
is a joint effort to extend mobile networks to support management system. The connection to the IoT
standardized, automatic device and subscription device itself, to the manufacturer or even to the
provisioning, it is at a very early stage. installer of the device can also be established.
ZTP support Operator
The main purpose of this service is to drive the application profile

A UICC MODULE CAN IoT device through the steps of automatic device
provisioning from the very beginning (ordering the
STORE SENSITIVE device) to the final decommissioning.
INFORMATION Although this over-the-top service (OTT) Figure 3 ZTP system with central ZTP service and UICC support
can speed up the provisioning process significantly,
During the provisioning procedure, two or more it has some disadvantages. It should not store sensitive
identities are given to the device, which entails data, but only manage it indirectly. Furthermore, This is where a UICC application can help and every use case and solve the problems the IoT area
that these identifiers are downloaded, and different if the device has no connection at all, it cannot support an OTT ZTP service. A UICC module can is struggling with today.
subsystems are configured (mobile network, device do anything. Scaling could also be a problem. store sensitive information from different security The UICC application can be used to monitor
domains. As it works close to the IoT device, it can do connectivity and fix issues locally. This can be
corrective actions locally if there is a problem with highly effective if credentials are stored on the
the connectivity (attempt to activate another profile UICC module and if the IoT protocol stack
and connect to another operator). In addition, it is is also running on the UICC module.
Terms and abbreviations scaling together with the IoT devices. Since this For narrowband IoT, the traditional profile
AKA – Authentication and Key Agreement |BIP – Bearer Independent Protocol | CoAp – Constrained solution is completely under the control of the download solution and the machine-to-machine
Application Protocol | DTLS – Datagram Transport Layer Security | eUICC – Embedded UICC (soldered to operator, it can be independent of the application, SM-DP is ineffective. Significantly better results
the device board) | HTTPS – Hypertext Transfer Protocol Secure | IMEI – International Mobile Equipment thereby also saving development costs. can be achieved by using the SM-DP+ in a new way.
Identity | IOT – Internet of Things | IUICC – Integrated UICC (integrated to a microchip) | LPA – Local Figure 3 shows an example of this system: For example, running the LPA proxy on the UICC
Profile Assistant | LPAdv – LPA (device), interfacing to the UICC | LPApr – LPA (proxy), interacting with the
a central ZTP service, in connection with module makes it possible to use completely new
device owner and SM-DP+ | LwM2M – Lightweight Machine-to-Machine | NIDD – Non-IP Data Delivery |
multiple subsystems and a support application options for device provisioning.
OMA – Open Mobile Alliance | OTT – Over-the-Top | PSK – Pre-shared Keys | RMS – Remote Management
Subsystem | RSP – Remote SIM Provisioning (protocol) | SCEF – Service Capability Exposure Functions |
on the UICC module.
The central ZTP service working together with Conclusion
SM-DP – Subscription Manager–Data Preparation | UICC – Universal Integrated Circuit Card |
USSD – Unstructured Supplementary Service Data | ZTP – Zero-Touch Provisioning the ZTP support application on the UICC module The universal integrated circuit card (UICC)
can be very effective. The ZTP service and the ZTP modules present in all 3GPP IoT devices
support application together can cover almost today are costly and underutilized.

26
8 ER
ERICS S IOCNS S
TEOCNHTNEO
CLHONGOYL O
RGE VYI ERW
E V✱
IEW
A P✱
R I L# 1042, 2 0 2 0 #P
A 2 I2L01240, 2 ✱
0R 0 2 0E R✱
ICS
ERS IOCNS S
TOE CNHTNEO
CLHONGOYL O
RGE VYI ERW
EVIEW 27
9
✱ UICC MODULES AND THE IoT UICC MODULES AND THE IoT ✱

The industry is looking for ways to replace them beyond the cellular domain. For example,
with a next-generation solution, but for the members of the existing UICC ecosystem can
foreseeable future UICC modules are here to stay. start exploiting UICC capabilities for storing
While there are a few ways to reduce the IoT identities, executing IoT protocols,
complexity of using UICC modules and thereby increasing security, providing end-to-end
reducing the cost of IoT devices, it is also possible connectivity as a service, and/or supporting
to extend the application of UICC modules well zero-touch provisioning.

References
1. Ericsson blog, Evolving SIM solutions for IoT, May 27, 2019, Smeets, B; Ståhl, P; Fornehed, J, available at:
https://www.ericsson.com/en/blog/2019/5/evolving-sim-solutions-for-iot
2. UICC card HW specification for P5Cxxxx cards, available at: http://www.e-scan.com/smart-card/nxp.pdf
3. GSMA, RSP Technical Specification Version 2.1, February 27, 2017, available at: Zsigmond Pap
https://www.gsma.com/newsroom/wp-content/uploads/SGP.22_v2.1.pdf ◆ joined Ericsson in 2012.

the authOrs
4. GSMA, Remote Provisioning Architecture for Embedded UICC Technical Specification Version 4.0, After working in the cloud
February 25, 2019, available at: https://www.gsma.com/newsroom/wp-content/uploads/SGP.02-v4.0.pdf native and 5G packet core
5. GSMA Intelligence: The future of the SIM: potential market and technology implications for the mobile areas as technical manager
ecosystem, February 2017, Iacopino, P; Rogers, M, available at: https://www.gsmaintelligence.com/ and system architect
research/?file=3f8f4057fdd7832b0b923cb051cb6e2c&download respectively, he moved into
6. OMA, Lightweight Machine to Machine Technical Specification: Core, July 10, 2018, available at: the IoT area. He specializes
http://www.openmobilealliance.org/release/LightweightM2M/V1_1-20180710-A/OMA-TS-LightweightM2M_ in low-level software
Core-V1_1-20180710-A.pdf Benedek Kovács development and he has Zsolt Vajta
◆ joined Ericsson in 2005. participated in multiple ◆ joined Ericsson in 2015
7. ARM, ARM Security Technology, available at: http://infocenter.arm.com/help/topic/com.arm.doc.prd29-
Over the years since he has hardware and firmware as a software developer
genc-009492c/PRD29-GENC-009492C_trustzone_security_whitepaper.pdf
served as a system engineer, developments related to focused on developing
8. GSMA, IoT SAFE, available at: https://www.gsma.com/iot/iot-safe/ R&D site innovation custom hardware solutions. and maintaining modules
9. OMA, white paper, Lightweight M2M 1.1: Managing Non-IP Devices in Cellular IoT Networks, October manager (Budapest) and He holds an M.Sc. in the area to implement the link
2018, Slovetskiy, S; Magadevan, P; Zhang, Y; Akhouri, S, available at: https://www.omaspecworks.org/wp- characteristics, performance of hardware and embedded aggregation control protocol
content/uploads/2018/10/Whitepaper-11.1.18.pdf management and reliability computers and a Ph.D. in the IP operating system.
The authors would
like to thank the specialist in the development in information engineering In 2018, he became involved
following people of the 4G VoLTE solution. from the Budapest University in research on IoT device
for their Today he works on 5G of Technology and activation and zero-touch
Further reading networks and distributed Economics in Hungary. provisioning. As of early
contributions
❭ Ericsson Technology Review, Key technology choices for optimal massive IoT devices, January 2019, to this article: cloud, as well as coordinating 2020, he has joined the
available at: https://www.ericsson.com/en/reports-and-papers/ericsson-technology-review/articles/key- Gergely Seres, global engineering projects. packet core area as a
technology-choices-for-optimal-massive-iot-devices John Fornehed, Kovács holds an M.Sc. in product owner. He holds
❭ Ericsson, eSIM – Let’s talk business, available at: https://www.ericsson.com/en/digital-services/trending/esim Per Ståhl, Peter information engineering and an M.Sc. in informatics and
Mattsson, Bogdan a Ph.D. in mathematics from physics as well as a Ph.D.
❭ Ericsson blog, Secure IoT identities, available at: https://www.ericsson.com/en/blog/2017/3/secure-iot-identities
Dragus, Robert the Budapest University of in nuclear physics from
❭ Ericsson blog, Secure brokering of digital identities, available at: https://www.ericsson.com/en/blog/2017/7/ Technology and Economics the University of Debrecen
Khello and
secure-brokering-of-digital-identities
Tony Uotila. in Hungary. in Hungary.

28
10 ERICS
ERS IOCNS S
TEOCNHTNEO
CLHONGOYL O
RGE VYI ERW
E V✱
IEW
A P✱
R I L# 1042, 2 0 2 0 #P
A 2 I2L01240, 2 ✱
0R 0 2 0E R✱
ICS
ERS IOCNS S
TOE CNHTNEO
CLHONGOYL O
RGE VYI ERW
EVIEW 29
11
✱ CTO TECHNOLOGY TRENDS 2020 CTO TECHNOLOGY TRENDS 2020 ✱

FUTURE NETWORK
TRENDS
CREATING INTELLIGENT DIGITAL INFRASTRUCTURE

BY: ERIK EKUDDEN, CTO

All around the world, the unprecedented that is always available to any consumer The machines and other ‘things’ that
events of 2020 have brought into focus of the digital infrastructure. The network make up the Internet of Things (IoT) require
the critical role that digital infrastructure platform forms the core of the digital even more sophisticated communication
plays in the functioning of virtually infrastructure, with the ability to ensure than humans do. For example, connected,
every aspect of contemporary society. long-term competitiveness for enterprises intelligent machines must be able to
More than ever before, communication and meet the full range of societal needs as interact dynamically with the network.
technologies are providing innovative well. It is a trustworthy solution that Sensor data will be used to support the
solutions to help address social, guarantees resilience, privacy, reliability development of pervasive cyber-physical
environmental and economic challenges and safety for all types of organizations – systems consisting of physical objects
by enhancing efficiency and enabling public, private and everything in between. connected to collaborative digital twins.
both intensified network usage and It also has the scale, cost performance and Future network capabilities will also include
more well-informed decisions. quality required to support future innovations. support for the transfer of sensing
As a result of these characteristics, it is the modalities such as sensations and smell.
One of the most important features of digital most sustainable solution to address all The network platform acts as a seamless
infrastructure is the ability to bridge distances future communication needs. universal connectivity fabric characterized
and make it easier to efficiently meet societal Future technologies will enable a fully by its almost limitless scalability and
needs in terms of resource utilization, digitalized, automated and programmable affordability. It is capable of exposing
collaboration, competence transfer, status world of connected humans, machines, capabilities beyond communication
verification, privacy protection, security and things and places. All experiences and services, such as embedded compute and
safety. The communications industry sensations will be transparent across the storage as well as a distributed intelligence
supports other industries by enabling them boundaries of physical and virtual realities. that supports users with insights and
to deliver digital products and services such Traffic in future networks will be generated reasoning.
as health care, education, finance, commerce, not only by human communication but also In this article, I will explain the ongoing
governance and agriculture. It also plays a by connected, intelligent machines and evolution of the network platform in terms
vital role in tackling climate change by helping bots that are embedded with artificial of the key needs that are driving its
other industries reduce emissions and intelligence (AI). As time goes on, the evolution (trends 1-3) and the emerging
improve efficiency. percentage of traffic generated by humans capabilities that will meet both those
In last year’s trends article, I introduced will drop as that of traffic generated by and other needs (trends 4-7).
the concept of the network platform and machines and computer vision systems –
explained how it serves as a catalyst in the including autonomous vehicles, drones
development of an open marketplace and surveillance systems – rises.

30 ERICSSON TECH NOLOGY REVIEW ✱ #02 2020 #02 2020 ✱ ERICSSON TECHNOLOGY REVIEW 31
✱ CTO TECHNOLOGY TRENDS 2020 CTO TECHNOLOGY TRENDS 2020 ✱

Trends 1-3: The key drivers The network platform will provide
an automated environment in which
the internet of senses, which combines
visual, audio, haptic and other technologies

of network platform evolution interconnected, intelligent machines

can communicate, including support for
AI-to-AI communication and autonomous
systems such as communication among
to allow human beings to have remote
sensory experiences.
The internet of senses will enable
seamless interaction with remote things
USE CASE

DIGITAL TWIN
The three key drivers that are most significant to the evolution of the network platform are
self-driving vehicles and intelligent
machines in factories.
and machines, making it possible to fully
realize use cases such as remote health
IN THE PORT
all related to bridging the gap between physical reality and the digital realm. Most notably, Intelligent machines have their own way
of perceiving information (data), which is
checks, remote operation of machinery,
holographic communication and virtual
OF LIVORNO
this involves delivering sensory experiences over networks and utilizing digital representations different from how humans perceive it. reality (VR) vacations. Among other benefits, Ericsson has deployed a digital twin
to make the physical world fully programmable. For example, communication among the internet of senses is expected to have a in the Italian port of Livorno (Leghorn).
intelligent machines requires new types of significant impact in terms of sustainability, As a result, terminal port operations
video compression mechanisms, as today’s by dramatically reducing the need for travel. will increasingly become a mixture
video codecs are optimized for human In the years ahead, major leaps forward of physical machinery, robotics
perception. are expected in sensor and actuator systems, automated vehicles,
Another aspect to consider is how technologies, such as the actuation of human-operated digital platforms
intelligent machines will interact and smell and high-quality touch sensation. and AI-based software systems.
communicate with each other. To improve During remote operations, the advances in All those components, served by
the reliability and efficiency of machine- haptic devices will allow virtual objects a 5G solution, transform the port
to-machine communication, machines will to be perceived just as the real objects environment into a ‘playground’
need to understand the meaning of the themselves. Holographic communication in which to experience the future
communication in terms of capabilities, will be possible without wearing extended of an automated physical world.
TREND #1: by physical objects can be used to create TREND #2: intentions and needs. This will require reality glasses, due to 3D light field display The port’s digital twin makes use
A COLLABORATIVE, AUTOMATED their digital twins. Collaborative digital CONNECTED, INTELLIGENT semantics-driven communication. technologies. of a plethora of real-time data
PHYSICAL WORLD twins will have the ability to manage the MACHINES Cognition is one of the most important Body augmentation capabilities will enable captured by connected objects at
As physical and digital realities become interactions between the physical objects Machines will become increasingly capabilities of an intelligent machine. humans to be smarter, stronger and more the physical port, including sensors,
increasingly interconnected, advanced they represent. intelligent and autonomous as their Cognitive machines are capable of capable. Other examples are contact lenses cameras and vehicles. An AI operation
cyber-physical systems have begun to Digitalizing the physical environment cognitive abilities continue to expand. self-learning from their interactions and that can display augmented reality (AR) management system operates on the
emerge. These systems consist of humans, in which the physical objects interact Their understanding of the world around experiences with their environment. content, universal translator earbuds digital model to determine the
physical objects (machines and other things), requires sensor data fusion – that is, them will continue to grow in tandem with They generate hypotheses and reasoned that allow for language-independent sequence of logistics tasks and
processes, networking and computation, using data from multiple sources to their ability to interact with other machines arguments, make recommendations and communication and exoskeletons that activities. Feedback from these
and the interactions between them all. create an accurate digital representation as part of a cognitive system of systems. take actions. They can adapt and make increase physical strength. Eventually, processes provides live updates
Their primary purpose is to provide individuals, of the physical environment. One example An intelligent machine uses sensors to sense of complexity and handle brain- computer interfaces will enable to the human supervisors using
organizations and enterprises with full of sensor data fusion is achieving high- monitor the environment and adjust its unpredictability. The future network will communication at the speed of thought VR and to the docks/quay
transparency to monitor and control assets precision positioning by combining actions to accomplish specific tasks empower cognitive machines by providing where, instead of speaking to machines, operators through AR.
and places, thereby generating massive network-based positioning data with in the face of uncertainty and variability. them with new network features and services humans will merely think in order to Results indicate that there are
benefits in terms of efficiency. One early information from other sensors such as These machines include three major such as sensing, high-precision positioning direct them. about 60 direct and indirect benefits
example of this is the way that cyber-physical cameras and inertial measurement units. subsystems: sensors, actuators and control. and distributed computing capabilities. The network platform supports the of the solution, including improved
systems can help planners optimize energy Ultimately, the joint communication Examples of intelligent machines include internet of senses with novel network competitiveness, increased safety
and materials usage. and sensing in future systems will make it industrial robots, speech recognition/ TREND #3: enablers such as distributed compute, high- for personnel, sustainable growth of
Soon, there will be hundreds of billions of possible to leverage all the interconnected voice synthesis and self-guided vehicles. THE INTERNET OF SENSES precision positioning, integrated sensing the port city, improved management
connected physical objects with embedded digital twins and digital representations The complexity of control and logic skills The ability to deliver multisensory experiences and application programming interfaces. of logistics and a positive
sensing, actuation and computing of the environment to create a complete makes expert systems central in the realm over future networks will make it easier than These are needed to support bandwidth environmental impact.
capabilities, which continuously generate digital representation of everything. of intelligent machines. ever before to transfer skills over the internet. and latency reservation, network latency
informative data. The sensor data generated It will ultimately lead to the emergence of reporting and network slice prioritization.

32 ERICSSON TECH NOLOGY REVIEW ✱ #02 2020 #02 2020 ✱ ERICSSON TECHNOLOGY REVIEW 33
✱ CTO TECHNOLOGY TRENDS 2020 CTO TECHNOLOGY TRENDS 2020 ✱

Trends 4-7: Critical enablers The design of terahertz electronics includes

very small antenna and radio frequency
fabric, connectivity, compute and storage
will be integrated, interacting to provide
Aside from the applications, the network
also provides a continuous execution

of the future network platform (RF) elements as well as high-performance

oscillators.
Full duplex is another component that can,
in some specific scenarios, substantially
maximum performance, reliability,
low jitter and millisecond latencies
for the applications they serve.
Rather than processing data centrally,
environment for access and core functions.
It runs on a distributed cloud infrastructure
with integrated acceleration for data-
intensive virtual network functions and
increase the link capacity compared with in many cases it is more efficient in terms applications.
The network platform is designed to deliver the kind of extreme performance required by half duplex. Full duplex is made possible by of bandwidth and/or latency constraints The future network platform goes
application areas such as the internet of senses and communication among intelligent machines. self-interference suppression circuits. to bring the processing closer to where beyond the use of microservices to
Visible light wireless communication, the data is produced, insights are consumed implement network functions as serverless
It will also serve new types of devices with close-to-zero-cost and close-to-zero-energy piggybacking on the wide adoption of LED and actions are taken. In some cases, local architectures. The server management and
implementations, which can be embedded into everything. Looking ahead, increasingly (light-emitting diode) lighting, is another operation may be required by regulations or capacity planning decisions are fully
advanced technologies in four areas (trends 4-7) will expand the capabilities of the digital potential step in the frequency domain to preferred for privacy, security or resilience autonomous from the developer and the
complement RF communications. reasons. network operator. The network takes care
infrastructure through the network platform.
Network as a sensor
Higher frequencies will further enhance the
spatial and temporal resolution of the radio
signal. Reflections of such radio signals can
be used to sense the surroundings.
Furthermore, high frequencies have
distinct atmospheric and material
interactions, where different frequencies
are more or less susceptible to things like
TREND #4: Flexible network topologies Access for zero-energy devices absorption in water, for example. This has
OMNIPRESENT AND NON- and deployments The rapidly growing demand for vast been shown to be sufficient to forecast
LIMITING CONNECTIVITY Network topologies and deployments will numbers of connected sensors and weather and air quality.
The concept of ubiquitous radio access is need to become increasingly flexible to actuators has made it necessary to invent Distance information to reflecting
evolving toward the vision of a future network provide coverage everywhere and deliver zero-energy devices. These will be deployed surfaces can be identified by integrating
that will deliver non-limiting performance extreme performance. One possibility is a once and will continuously report and act positioning and sensing capabilities.
to satisfy the needs of humans, things and multi-hop-based radio network, where a without the need for maintenance or Such information can be used to detect
machines by enhancing multidimensional multitude of nodes collaborate to forward external charging. The stepping stones obstacles and speed as well as to generate
coverage, stellar capacity and augmenting a message to the receiver. This solution is along the way include narrowband IoT real-time local maps.
capabilities. particularly interesting for smaller cells enhancements and massive machine
of limited reach. Satellites, high-altitude type communication for 5G New Radio TREND #5:
Access coverage everywhere platforms and airborne cells can be for local area networks (LANs) as well as PERVASIVE NETWORK
Further densification of networks is needed integrated into the network as a complement for wide-area usage. COMPUTE FABRIC
to provide high-speed coverage everywhere. to extend coverage. Further components in As distributed compute and storage
Connected airborne devices, such as drones, a flexible topology can include connected Extreme radio performance continues to evolve, the lines between
require access on altitudes up to several device relay and the possibility for ad-hoc The network will utilize higher frequency the device, the edge of the network and
kilometers, making it necessary to have a deployments of networks. Ultimately, bands to deliver extreme performance. the cloud will become increasingly blurred.
3D point of view including the elevation distributed massive MIMO (multiple-input, For example, communications over the Everything can be viewed as a single,
aspect to provide coverage. There is also multiple-output) solutions may lead to fully terahertz frequency band (above 100GHz) unified, integrated execution environment
a need to ensure high-performing indoor distributed connectivity, where many radio have some attractive properties, for distributed applications, including
connectivity by increasing the number of network nodes simultaneously serve a user, including terabit-per-second link both network functions and third-party
indoor small cells and fully integrating them. without fixed-cell borders. capacities and miniature transceivers. applications. In the network compute

34 ERICSSON TECH NOLOGY REVIEW ✱ #02 2020 #02 2020 ✱ ERICSSON TECHNOLOGY REVIEW 35
✱ CTO TECHNOLOGY TRENDS 2020 CTO TECHNOLOGY TRENDS 2020 ✱

of the deployment, scaling and all resources These identities are securely anchored radio interface, automation of network
required to ensure that the function to devices and network nodes by root- management and orchestration such as
deployed is always available at any scale. of-trust mechanisms. the optimization of parameters, handling of
Upcoming novel computing architectures Network platform solutions utilize alarms and self-healing. AI algorithms will
include memory-centric computing, optical confidential computing to protect identities be deployed and trained at different
computing, nanocomputing, neuromorphic and their data and establish trust among network domains, for example, in
computing and even quantum computing. network customers and their assets, management, the core network and the
In the future, these architectures will enable thereby also offering assurance to users radio network. Physical layer algorithms,
continued exponential growth in compute and regulators. This requires automated such as link adaptation, handover, power
capacity for most applications running on trust assessment of all network elements, control and dynamic scheduling of
the network compute fabric – an important things, machines and applications, as well resources can be optimized with AI agents.
development as the end of Moore’s law as compute and storage resources by Network management will become less
approaches. using remote attestation and AI. complex through intelligent closed-loop
Responsible AI will bring trustworthy automation with support for humans to
TREND #6: automated protection and risk management. interact with the network and monitor its
TRUSTWORTHY AI-based automation provides the ability behaviors. The network operator expresses
INFRASTRUCTURE to act on a high number of events affecting the intent of a desired network state or goal,
Governments and enterprises are adopting the network infrastructure or the network and the network internally resolves the
advanced technologies for secure assurance usage. detailed steps necessary to achieve that
of mission- and business-critical processes intent. Network knowledge, data and
such as factory automation, remote control TREND #7: actions are shaped in such a way that the
of assets and more. The highly trustworthy COGNITIVE NETWORK operator interacting with the network can
network platform fulfills the requirements In the vision of zero-touch network understand them.
of even the most mission- and business- management and operations, networks The cognitive network will be based on
critical use cases. It offers a fusion of are deployed and operated with minimum control design, using both machine
connectivity and compute characterized by human intervention, using trustworthy reasoning and machine learning techniques
different dimensions of resilience, privacy, AI technologies. All operational processes that are distributed and capable of acting in
security, reliability and safety. It will also and tasks, including, for example, delivery, real time. The network is a highly distributed
provide adaptable and verifiable dimensions deployment, configuration, assurance system where multiple AI agents, present
of trustworthiness in a scalable and cost- and optimization, will be executed with across the network, need to interwork to
efficient manner. 100 percent automation. optimize overall network performance.
Rather than being designed per node The network itself will continuously Local decisions need to be coordinated
or for a particular part of the network, the learn from its environment observations, with more central intent-based decisions.
always-on characteristics of the network interactions with humans and previous The central AI agent needs to make decisions
platform such as reliability, availability and experiences. The cognitive processes in real time based on both local and global
resilience rise up to cover the complete understand the current network situation, information. Multiple distributed AI agents
network. Always-on mechanisms are built plan for wanted outcome, decide on what share distributed insights throughout
into user plane, control plane and device to do and act accordingly. The outcome the network through federated learning.
mobility solutions. All parts of the network serves as an input to learn from its actions. Cognitive networks will be inherently
will be addressed including transport The cognitive network will be able to trustworthy – that is, reliable, safe,
nodes and transport networks, network optimize its existing knowledge, build on secure, fair, transparent, sustainable
infrastructure and site solutions. experience and reason in order to solve and resilient – by design.
To protect communication and data, new problems.
secure identities are utilized at every The network will utilize intent-based
layer between humans, devices and and distributed intelligence for multiple
applications in different industry segments. functions, including optimization of the

36 ERICSSON TECH NOLOGY REVIEW ✱ #02 2020 #02 2020 ✱ ERICSSON TECHNOLOGY REVIEW 37
✱ CTO TECHNOLOGY TRENDS 2020 CTO TECHNOLOGY TRENDS 2020 ✱

CONCLUSION

The network platform is

the spinal cord of intelligent
digital infrastructure
The digital infrastructure offers endless It is also the ideal platform for all types of the growth and expansion of intelligent
possibilities to individuals, enterprises innovation, with the ability to support digital infrastructure on the network platform.
and governments across the globe, interactions that empower an intelligent, The first three of the seven trends this year
with its unique ability to bridge vast sustainable and connected world. are the key drivers of network platform
distances and enable powerful new The major advantage of the network evolution – the creation of a collaborative
solutions to a wide range of social, platform is that it will be accessible automated physical world, connected,
environmental and economic anywhere, always-on and with guaranteed intelligent machines and the internet of
challenges. Health care, education, performance. Nomadic distributed senses. All three highlight the growing need
finance, commerce, governance and processing and storage will be embedded to bridge the gap between physical and
agriculture are just a few of the sectors into it to support advanced applications. digital realities. Trends 4-7 are increasingly
that stand to benefit from the massive It will be inherently reliable and resilient, advanced technologies in four areas –
efficiency gains that digital fulfilling all the requirements for secure non-limiting connectivity, pervasive
infrastructure can provide. communication. Cognitive operations network compute fabric, trustworthy
and maintenance of the network and its infrastructure and cognitive networks.
Designed to carry vital messages, services will deliver the most cost-efficient Breakthroughs in these four areas will be
commands, reasoning, insights, intelligence and sustainable solution to meet any essential to fully enable trends 1-3 and
and all the sensory information needed to and all communication needs. continuously expand the capabilities of the
support the continuous evolution of industry With this in mind, it is clear that the most digital infrastructure through the network
and society, the network platform is designed important future network trends to watch in platform in the years and decades ahead.
to be the spinal cord of digital infrastructure. 2020 are those that relate most closely to

◆ As Group CTO, Erik Ekudden is responsible for setting the direction of technology leadership
for the Ericsson Group. His experience of working with technology leadership globally influences Further reading
the strategic decisions and investments in, for example, mobility, distributed cloud, artificial intelligence ❭ Ericsson blog, What do cyber-physical systems have in store for us?, available at: https://www.ericsson.com/en/blog/2019/12/
and the Internet of Things. This builds on his decades-long career in technology strategies and industry cyber-physical-systems-technology-trend
activities. Ekudden joined Ericsson in 1993 and has held various management positions in the company, ❭ Ericsson report, 10 Hot Consumer Trends 2030, available at: https://www.ericsson.com/en/reports-and-papers/consumerlab/
including Head of Technology Strategy, Chief Technology Officer Americas in Santa Clara (USA), reports/10-hot-consumer-trends-2030
and Head of Standardization and Industry. He is also a member of the Royal Swedish Academy
❭ Ericsson blog, Driving business value in an open world, available at: https://www.ericsson.com/en/blog/2020/7/cto-driving-business-
of Engineering Sciences and the publisher of Ericsson Technology Review.
value-in-an-open-world
❭ Ericsson Technology Review, CTO Technology Trends 2019, available at: https://www.ericsson.com/en/reports-and-papers/
ERIK EKUDDEN
ericsson-technology-review/articles/technology-trends-2019
SENIOR VICE PRESIDENT, CHIEF TECHNOLOGY OFFICER
AND HEAD OF GROUP FUNCTION TECHNOLOGY

38 ERICSSON TECH NOLOGY REVIEW ✱ #02 2020 #02 2020 ✱ ERICSSON TECHNOLOGY REVIEW 39
✱ THE FUTURE OF CLOUD COMPUTING THE FUTURE OF CLOUD COMPUTING ✱

The future of
The result is that cloud computing resources Figure 1 positions the four key challenges in
are becoming increasingly heterogeneous, while relation to the orchestration/operations support
simultaneously being widely distributed across systems (OSS) layer, the application layer, run-time
smaller DCs at multiple locations. Cloud deployments and the operating system/hypervisor. The lower part
must be rethought to address the complexity and of the figure provides some examples of specialized
technical challenges that result from this profound HW in a telco environment, which includes domain-

cloud computing
transformation. specific accelerators, next-generation memory and
In the context of telecommunication networks, storage, and novel interconnect technologies.
the key challenges are in the following areas:
Compute and storage trends
1. Virtualization of specialized HW resources With the inevitable end of Moore’s Law [2],
2. Exposure of heterogeneous HW capabilities developers can no longer assume that rapidly
3. HW-aware workload placement increasing application resource demands
4. Managing increased complexity. will be addressed by the next generation
HIGHLY DISTRIBUTED WITH HETEROGENEOUS HARDWARE of faster general-purpose chips. Instead,
Getting all these pieces right will enable the commodity HW is being augmented by a very
future network platform to deliver optimal heterogeneous set of specialized chipsets,
application performance by leveraging emerging referred to as domain-specific accelerators,
HW innovation that is intelligently distributed that attempt to provide both cost and
With a vastly distributed system (the telco network) already in place, throughout the network, while continuing to energy savings.
the telecom industry has a significant advantage in the transition harvest the operational and business benefits For instance, data-intensive applications can take
of cloud computing models. advantage of the massive scope for parallelization
toward distributed cloud computing. To deliver best-in-class application
performance, however, operators must also have the ability to fully
leverage heterogeneous compute and storage capabilities. HW-aware
Orchestration/OSS Managing increased
workload placement complexity

Exposure of
HW capabilities

Application Application 5G UPF 5G gNB

The cloud is transforming, both in terms of cases have matured and new ones have continued Run-time Virtualization of Run-time User
WOLFGANG JOHN,
device
CHANDR AMOULI the extent of distribution and in the diversity of to emerge, requirements on latency, energy Operating system/hypervisor
specialized HW Operating system/hypervisor
SARGOR , ROBERT compute and storage capabilities. On-premises efficiency, privacy and resiliency have become
SZABO, AHSAN and edge data centers (DCs) are emerging, more stringent, while demand for massive data Distributed compute Distributed compute
JAVED AWAN, CHAKRI and hardware (HW) accelerators are becoming storage has increased. & storage HW & storage HW
PADALA , EDVARD integral components of formerly software-only To meet performance, cost and/or legal • Memory pooling Next-generation • Near-memory computing
DR AKE , MARTIN services. requirements, cloud resources are moving toward • Storage-class memories memory & storage • PMEM
JULIEN, MILJENKO the edge of the network to bridge the gap between • GPUs/TPUs Domain-specific • GPUs/ASICs
OPSENICA ■ One of the main drivers into the age of resource-constrained devices and distant but • FPGAs accelerators • FPGAs and SmartNICs

virtualization and cloud was the promise of powerful cloud DCs. Meanwhile, traditional COTS • Cache-coherent interconnects Novel interconnect • Cache-coherent interconnects
• High bandwidth interconnects technologies • High bandwidth interconnects
reducing costs by running all types of workloads HW is being augmented by specialized
on homogeneous, generic, commercial off-the- programmable HW resources to satisfy the strict Central Edge
shelf (COTS) HW hosted in dedicated, performance requirements of certain applications
centralized DCs. Over the years, however, as use and limited energy budgets of remote sites. Figure 1 Impact of the four key challenges on the stack (top) and heterogeneity of HW infrastructure (bottom)

40
2 E R I ECRSISCOSNS O
TNE CTHEN
COH LNOOGL Y
OGRY
E VRI E V
WI E✱W M
✱AY# 1022, 2 0 2 0 2 21 20 ,2200 2✱
# 0AY
M 0 E✱
R I ECRSISCOSNS O
TNE CTHEN
COH LNOOGL Y
OGRY
E VRI E V
WI E W 41
3
✱ THE FUTURE OF CLOUD COMPUTING THE FUTURE OF CLOUD COMPUTING ✱

in graphics processing units (GPUs) or tensor Supporting heterogeneous hardware physical accelerator into multiple virtual accelerators of resources and capabilities. During exposure and
processing units (TPUs), while latency-sensitive in distributed cloud must be done manually. Addressing these issues abstraction, proprietary features and the interfaces
applications or locations with limited power budgets While the combination of heterogeneous HW will require appropriate abstractions and models to them must be hidden and mapped to (formal or
may utilize field-programmable gate arrays and distributed compute resources poses unique of specialized HW, so that their capabilities can be informal) industry standards that are hopefully
(FPGAs). These trends point to a rapidly increasing challenges, there are mechanisms to address interpreted and incorporated by orchestration coming soon. Modeling and abstraction of resources
adoption of accelerators in the near future. each of them. functions. and capabilities are necessary prerequisites to be
The growing demand for memory capacity from The need for appropriate models will be further able to select the appropriate location and
emerging data-intensive applications must be met by Virtualization of specialized hardware amplified in the case of distributed compute and application deployment options and flavors.
upcoming generations of memory. Next-generation The adoption of specialized HW in the cloud storage. Here, the selection of the optimal site
memories aim to blur the strict dichotomy between enables multiple tenants to use the same HW location will depend on the application requirements Orchestrating heterogeneous distributed cloud
classical volatile and persistent storage technologies – under the illusion that they are the sole user, (bounded latency or throughput constraints, for Based on a global view of the resources and
offering the capacity and persistence features of with no data leakage between them. The tenants example) and the available resources and HW capabilities within the distributed environment,
storage, combined with the byte-addressability can request, utilize and release accelerators at any capabilities at the sites. The programming and an orchestration system (OSS in telco terminology)
and access speeds close to today’s random-access time using application programming interfaces orchestration models must be able to select typically takes care of designing and assigning
memory (RAM) technologies. Such persistent (APIs). This arrangement requires an abstraction appropriate software (SW) options – SW only in the application workloads within the compute and
memory (PMEM) technologies [3] can be used layer that provides a mechanism to schedule jobs case of moderate requirements, for example, or SW storage of the distributed environment. This
either as large terabyte scale volatile memory, or as to the specialized HW, monitor their resource complemented with HW acceleration for stringent means that decisions regarding optimal workload
storage with better latency and bandwidth relative usage and dynamically scale resource allocations requirements. placement also should factor in the type of HW
to solid-state disks. to meet performance requirements. It is pertinent As SW deployment options with or without HW components available at the sites related to the
to keep the overhead of this virtualization to a acceleration may have significantly different requirements of the specific application SW.
minimum. While virtualization techniques for resource footprints, sites must expose their HW Due to the pricing of and power constraints
THESE TRENDS POINT TO A common COTS HW (x86-based central capabilities to be able to construct a topology map on existing and upcoming HW accelerators,
RAPIDLY INCREASING ADOPTION processing units (CPUs), dynamic RAM (DRAM),
block storage and so on) have matured well during
OF ACCELERATORS... recent decades, corresponding virtualization
techniques for domain-specific accelerators are Definition of key terms
3D silicon die-stacking has facilitated the largely still missing for production-grade systems. Edge computing provides distributed computing and storage resources closer to the location where they
embedding of compute units directly inside memory are needed/consumed.
and storage fabrics, opening a paradigm of near- Exposure of hardware capabilities
memory processing [1], a technology that reduces Current cloud architectures are largely agnostic Distributed cloud provides an execution environment for cloud application optimization across multiple
sites, including required connectivity in between, managed as one solution and perceived as such by the
data transfer between compute and storage and to the capabilities of specialized HW. For example,
applications.
improves performance and energy consumption. all GPUs of a certain vendor are treated as
Finally, advancements in interconnect technologies equivalent, regardless of their exact type or make. Hardware accelerators are devices that provide several orders of magnitude more efficiency/
will enable faster speeds, higher capacity and lower To differentiate them, operators typically tag the performance compared with software running on general purpose central processing units for selected
latency/jitter to support communication between the nodes equipped with different accelerators with functions. Different hardware accelerators may be needed for acceleration of different functions.
various memory and processing resources within unique tags and the users request resources with
nodes as well as within clusters. The cache coherency a specific tag. This model is very different to Persistent memory is an emerging memory technology offering capacity and persistence features of
block-addressable storage, combined with the byte-addressability and access speeds close to today’s
properties of modern interconnect technologies, general-purpose CPUs and can therefore lead to
random-access memory technologies. It is also referred to as storage-class memory.
such as Compute Express Link [4] and Gen-Z, can complications when a user requires combinations
enable direct access to configuration registers and of accelerators. Moore's law holds that the number of transistors in a densely integrated circuit doubles about every two
memory regions across the compute infrastructure. Current deployment specifications also do not years, increasing the computational performance of applications without the need for software redesign.
This will simplify the programmability of accelerators have good support for requesting partial allocation Since 2010, however, physical constraints have made the reduction in transistor size increasingly difficult
and facilitate fine-grained data sharing among of accelerators. For accelerators that can be and expensive.
heterogeneous HW. partitioned today, the decomposition of a single

42
4 E R I ECRSISCOSNS O
TNE CTHEN
COH LNOOGL Y
OGRY
E VRI E V
WI E✱W M
✱AY# 1022, 2 0 2 0 2 21 20 ,2200 2✱
# 0AY
M 0 E✱
R I ECRSISCOSNS O
TNE CTHEN
COH LNOOGL Y
OGRY
E VRI E V
WI E W 43
5
✱ THE FUTURE OF CLOUD COMPUTING THE FUTURE OF CLOUD COMPUTING ✱

When a service request arrives, the orchestration training of machine learning (ML) models on live
HW capability
Business
service designs the service instance topology and network data. The enhancements to interconnects
exposures Zero-touch orchestration
Network slices intent assigns resources to each service component (cache coherency, for example) make it easier for the
instance (red arrows). These actions are based on various accelerators and CPUs to work together.
Gaming HW capability
AR/VRB APP APP control the actual service requirements, the service access The interconnects also enable low latencies and
points and the business intent. high bandwidths within sites and nodes. There is
E-MBB xNF xNF
increasing demand on memory from several core
Opportunities and use cases network functions (user-database functions,
APP In terms of the opportunities in support of the for example), both from a scale and a latency
Automotive xNF ongoing cloudification of telco networks, let us perspective. The scale of PMEM can be intelligently
xNF APP
APP APP APP
consider the case of RAN. The functional split combined with the low latency of double data rate
SmartNICs
of higher and lower layers of the RAN protocol memories to address these requirements.
Manufacturing xNF
APP makes it possible to utilize Network Functions While these opportunities are specific to
xNF Public clouds
Internet of xNF Virtualization (NFV) and distributed compute telecommunication providers, there are also several
Things PMEM APP xNF infrastructure to achieve ease of deployment and classes of third-party applications that would benefit
xNF Central sites
management. The asynchronous functions in the from distributed compute and storage capabilities
Distributed sites higher layer may be able to be run on COTS HW. within the telco infrastructure. Industry 4.0 includes
Fixed access (edge/regional cloud) xNF: telco Virtual Network Function or However, a set of specialized HW will be required several use cases that could utilize HW-optimized
Access sites (edge cloud) Cloud-native Network Function
APP: Third-party application to meet the stringent performance criteria of lower- processing. Indoor positioning typically requires the
layer RAN functions. For instance, the time- processing of high-resolution images to accurately
synchronous functions in the medium-access determine the location of an object relative to others
Figure 2 Integrated network slicing (telco) and third-party applications
control layer, such as scheduling, link adaptation, on a factory floor. This is computationally intensive
power control, or interference coordination, typically and GPUs/FPGAs are typically used. Likewise,
they are expected to be scarce among edge-cloud from human-guided automation to machine- require high data rates on their interfaces that scale the application of augmented reality (AR)/virtual
sites, which in turn will require mechanisms to reasoning-based automation such as cognitive with the traffic, signal bandwidth and number of reality (VR) technologies in smart manufacturing
employ prioritization and preemption of workloads. artificial intelligence (AI) technologies. antennas. These cannot be easily met with current for remote assistance, training or maintenance
Unlike conventional IT cloud environments, Specifically, a paradigm is emerging where the general-purpose processing capabilities. will rely significantly on HW acceleration and
distributed cloud allows considerations of remote human input to the cloud system will be limited Likewise, deciphering functions in the packet edge computing to optimize performance and
resources and capabilities. to specifying the desired business objectives data convergence protocol layer, compression/ reduce latencies.
Moreover, telco applications and workloads (intents). The cloud system then figures how best decompression schemes of fronthaul links and The gaming industry is also witnessing
hosted in telco clouds may require much stricter to realize those objectives/intents. channel decoding and modulation functions in the significant technology shifts – specifically, remote
Service Level Agreements (SLAs) to be fulfilled. Figure 2 presents an exemplary distributed cloud physical layer would all benefit from HW rendering and mixed-reality technologies will have
Prioritization and preemption for new workloads scenario with access sites, regional and central DCs acceleration. a profound impact on the consumer experience.
may only be a viable option if capabilities or and public clouds. It is based on the assumption that The security requirements for data flows across These technologies rely on an underlying distributed
resources are already taken. However, it is important the manufacturing network slice (red) includes both the backhaul for 4G/5G RANs mandate the use of cloud infrastructure that has HW acceleration
to migrate evicted workloads either to a new telco (xNF) and third-party workloads (APP), IP security protocols (IPsec). By offloading encrypt/ capabilities at the edge to offload the processing
location, or to a new SW and HW deployment option out of which one APP requires network acceleration decrypt functions to specialized HW such as from consumer devices, while maintaining strict
to minimize service disruption during preemption. (SmartNIC), while another xNF depends on PMEM. Smart Network Interface Controllers (SmartNICs), latency bounds.
Multiple network slices are created based on application-specific integrated circuits (ASICs) Furthermore, several use cases in the automotive
Managing increased complexity customer need. Network slices differ not only in their or FPGAs, the processing overhead associated with industry involve strict latency requirements that
Traditional automation techniques based on service characteristics, but are separated and IPsec can be minimized. This is crucial to support demand HW acceleration in the form of GPUs and
human scripting and/or rule books cannot scale to isolated from each other. Aggregated views of higher data rates in the transport network. FPGAs at remote sites. Examples include real-time
address the complexity of the heterogeneous HW accelerators per location are collected for the The network data analytics function in 5G Core object detection in video streams that are processed
distributed cloud. We can already see a shift away zero-touch orchestration service (gray arrows). networks would benefit from GPUs to accelerate by either vehicles or road-side infrastructure.

44
6 E R I ECRSISCOSNS O
TNE CTHEN
COH LNOOGL Y
OGRY
E VRI E V
WI E✱W M
✱AY# 1022, 2 0 2 0 2 21 20 ,2200 2✱
# 0AY
M 0 E✱
R I ECRSISCOSNS O
TNE CTHEN
COH LNOOGL Y
OGRY
E VRI E V
WI E W 45
7
✱ THE FUTURE OF CLOUD COMPUTING THE FUTURE OF CLOUD COMPUTING ✱

through our HW management layer, comprising of (FPGA-based SmartNICs in this case), we achieve
Kubernetes cluster node x PCI bandwidth and memory-sharing functions (the the highest possible network input/output
User space Kernel orange boxes in the FPGA part of the figure). performance in a virtualized environment and
Kubelet
An HW management layer uses partial free up CPUs for the execution of other functions
reconfiguration technology to support spatial and applications.

Streaming
Direct memory access (DMA) driver

interface
FPGA exposure plugin
LDPC encoder partitioning of one FPGA to share its resources. It is
FPGA manager comprised of a bandwidth-sharing layer for dynamic Orchestrating heterogeneous distributed cloud

Bandwidth-sharing layer

Memory-sharing layer

Memory infrastructure
API PCIe
allocation of PCIe bandwidth. The memory-sharing There is a need for intelligent workload placement

DMA engine
layer protects tenants from data leaks on the off-chip schemes to meet the diverse acceleration needs

Streaming
interface
LDPC decoder
DRAM. It adds a protection layer for internal of 5G use cases that require domain-specific HW.
LDPC LDPC
encoder decoder reconfiguration to prevent unintended We have two proof points related to this issue.
configurations. This solution has been validated The first is our development of an HW-aware

Streaming
interface
ML ML inference using multiple regions, hosting 5G low-density service instance design and workload placement to
inference
Config. parity-check (LDPC) code encoder and decoder optimize the deployment of workloads at the edge.
control
Pod
accelerators next to an ML inference function. We have demonstrated an on-demand service
Host FPGA board Memory However, essentially any telco or third-party instance design, prioritization and preemption for
application accelerator could be deployed within the telco Packet Core Gateway (PCG) user-plane
Figure 3 FPGA sharing solution these partial regions, whose number could vary function (UPF) over a Kubernetes edge cluster, in
depending on the FPGA size and the application’s which only some nodes are equipped with PMEM.
resource requirements. The PCG-UPF is considered a high-priority
Initial proof points As high-end GPUs tend to be expensive and may Virtual switches form an integral part of workload, which could use PMEM to distribute
Our ongoing research in the area of distributed not be available in every node of the distributed distributed cloud infrastructure, providing network database instances needed for resiliency.
cloud has yielded several initial proof points that cloud, Ericsson Research has also devised a solution access to virtual machines by linking the virtual and At the instantiation time of the PCG-UPF,
demonstrate Ericsson’s leadership in terms of fully to enable GPU access by applications hosted on physical network interfaces. The overhead of these another workload using the PMEM was evicted.
leveraging heterogeneous compute and storage nodes without local GPUs, both enabled by virtual switches is one of the main obstacles to The challenge was to migrate the lower priority
capabilities to deliver best-in-class application OpenStack and Kubernetes. Given a high-speed achieving high throughput in the packet-processing workload to a new host, considering its original
performance to our customers. network between the nodes, GPU requests are functions. Through our solution to off-load the data service requirements. Here, the lower priority
locally intercepted and redirected to the remote plane of a virtual switch onto the specialized HW workload shared an affinity with other components,
Virtualized heterogeneous hardware GPU servers for execution.
The 3GPP network evolution requires a dramatic FPGAs may be another alternative to meet the
increase in compute capacity when increasing requirements of RAN functions and third-party
carrier bandwidths up to terahertz level and the applications hosted at the telco edge. At Ericsson
Terms and abbreviations
addition of more MIMO (multiple-input, multiple- Research, we have enabled multi-tenancy support
output) layers and antenna ports. High-end GPUs on FPGAs through Kubernetes, so that multiple- AI – Artificial Intelligence | API – Application Programming Interface | AR – Augmented Reality |
ASIC – Application-Specific Integrated Circuit | COTS – Commercial Off-the-Shelf | CPU – Central
could provide the required compute capacity. application containers that need HW acceleration
Processing Unit | DC – Data Center | DRAM – Dynamic Random-Access Memory | E-WBB – Enhanced
To understand how 5G baseband can be can share the FPGA’s internal resources, off-chip
Mobile Broadband | FPGA – Field-Programmable Gate Array | GNB – GNodeB (3GPP next-generation
implemented on GPUs, Ericsson has entered DRAM and peripheral component interconnect
base station) | GPU – Graphics Processing Unit | HW – Hardware | IPsec – IP Security Protocols |
into a collaboration with NVIDIA. Early findings express (PCIe) bandwidth, as shown in Figure 3. LDPC – Low-Density Parity-Check | ML – Machine Learning | NFV – Network Functions Virtualization |
show that the GPU instruction set and CUDA On the left side, three application containers OSS – Operations Support Systems | PCG – Packet Core Gateway | PCIe – Peripheral Component
(Compute Unified Device Architecture) (pods) on a node are managed through Kubernetes. Interconnect Express | PMEM – Persistent Memory | RAM – Random-Access Memory | SLA – Service
SW design patterns for key receiver algorithms Our FPGA exposure plugin supports the pods by Level Agreement | SmartNIC – Smart Network Interface Controller | TPU – Tensor Processing Unit |
excel in comparison with CPU-only solutions, offering three isolated, virtual FPGAs. On the right UPF – User-Plane Function | VR – Virtual Reality | xNF – network functions, both telco and third party
indicating that further investigation of GPUs side, three separate FPGA regions (one per
for this purpose is indeed a viable direction. application container) are managed and exposed

46
8 E R I ECRSISCOSNS O
TNE CTHEN
COH LNOOGL Y
OGRY
E VRI E V
WI E✱W M
✱AY# 1022, 2 0 2 0 2 21 20 ,2200 2✱
# 0AY
M 0 E✱
R I ECRSISCOSNS O
TNE CTHEN
COH LNOOGL Y
OGRY
E VRI E V
WI E W 47
9
✱ THE FUTURE OF CLOUD COMPUTING THE FUTURE OF CLOUD COMPUTING ✱

which were automatically migrated together is still optimal, and searches for improved
with the workload blocking the PMEM needed configurations and uses predictive models to drive
by the PCG-UPF. The service disruption time proactive actions to be taken, thereby enabling
for the evicted traffic was minimized with the intelligent autonomous cloud operations.
instant triggering of the migration workflow.
Our second proof point demonstrates how Conclusion
References
we can enable distributed applications to utilize The cloud is becoming more and more distributed
the enhanced capacity and persistency of non- at the same time that compute and storage 1. ArXiv, Near-Memory Computing: Past, Present, and Future, August 7, 2019, Gagandeep Singh et al.,
volatile memories in a transparent fashion. capabilities are becoming increasingly diverse. available at: https://arxiv.org/pdf/1908.02640.pdf
While PMEM could be used to make up for Datacenters are emerging at the edges of telco 2. Nature, The chips are down for Moore’s law, February 9, 2016, M. Mitchell Waldrop, available at: https://
the slow growth of DRAM capacity in recent networks and on customer premises and hardware www.nature.com/news/the-chips-are-down-for-moore-s-law-1.19338
years, it can lead to performance degradation accelerators are becoming essential components 3. Admin magazine, How Persistent Memory Will Change Computing, Jeff Layton, available at: https://www.
due to PMEM’s slightly higher latencies. of formerly software-only services. Due to the admin-magazine.com/HPC/Articles/Persistent-Memory
The memory-tiering concept developed by inevitable end of Moore’s law, the importance 4. CXL, Compute Express Link, Breakthrough CPU-to-device interconnect, available at: https://www.
Ericsson Research enables the dynamic placement/ and use of hardware accelerators will only computeexpresslink.org/
migration of application data across DRAM and continue to increase, presenting a significant
5. Ericsson, Network compute fabric, available at: https://www.ericsson.com/en/future-technologies/network-
PMEM, based on observed application behavior. challenge to existing solutions for exposure
compute-fabric
This infrastructure, using low-level CPU and orchestration.
performance counters to drive placement decisions, To address these challenges, Ericsson is
achieves performance similar to DRAM, without innovating in three key areas. Firstly, we are using
any changes to the application, while using a virtualization techniques for domain-specific Further reading
mix of PMEM and DRAM. accelerators to support sharing and multi-tenant ❭ Ericsson blog, How will distributed compute and storage improve future networks, available at: https://
use of specific hardware. Secondly, we are using www.ericsson.com/en/blog/2020/2/distributed-compute-and-storage-technology-trend
... PAVING THE WAY TOWARD zero-touch orchestration for hardware accelerators,
which includes hardware capability exposure and
❭ Ericsson white paper, Edge computing and deployment strategies for communication service providers,
AN INTEGRATED NETWORK aggregation for the orchestration system, as well as
available at: https://www.ericsson.com/en/reports-and-papers/white-papers/edge-computing-and-deployment-
strategies-for-communication-service-providers
COMPUTE FABRIC THAT IS automated mechanisms to design and assign service
instances based on the abstract map of resources ❭ Ericsson blog, Cloud evolution: the era of intent-aware clouds, available at: https://www.ericsson.com/en/
UNIVERSALLY AVAILABLE and accelerator capabilities. And thirdly, we are blog/2019/5/cloud-evolution-the-era-of-intent-aware-clouds
using artificial intelligence and cognitive ❭ Ericsson blog, What is network slicing?, available at: https://www.ericsson.com/en/blog/2018/1/what-is-
Complexity management technologies to address the technical complexity network-slicing
A heterogeneous and distributed cloud implies and to optimize for business value. ❭ Ericsson blog, Virtualized 5G RAN: why, when and how?, available at: https://www.ericsson.com/en/
high complexity in service assurance, and more Redefining cloud to expose and optimize the use blog/2020/2/virtualized-5g-ran-why-when-and-how
specifically, high complexity in terms of of heterogeneous resources is not straightforward, ❭ Ericsson Technology Review, Cognitive technologies in network and business automation, available at:
continually finding optimal configurations in and to some extent goes against the centralization https://www.ericsson.com/en/reports-and-papers/ericsson-technology-review/articles/cognitive-technologies-in-
dynamically changing environments. Ericsson’s and homogenization trends. However, we believe network-and-business-automation
cognitive layer has demonstrated cloud service that our use cases and proof points validate our
❭ Ericsson, A guide to 5G network security, available at: https://www.ericsson.com/en/
assurance while satisfying contracted SLAs. approach and will gain traction both in the
security/a-guide-to-5g-network-security
This cognitive process evaluates the effect of a telecommunications community and beyond,
proposed new service deployment on all existing paving the way toward an integrated network ❭ Ericsson, 5G Core, available at: https://www.ericsson.com/en/digital-services/offerings/core-network/5g-core
services and their SLA fulfilment. Furthermore, compute fabric [5] that is universally available ❭ Ericsson, Edge computing, available at: https://www.ericsson.com/en/digital-services/trending/edge-
the cognitive layer continuously reevaluates across telco networks. computing
whether the current deployment of all services

48
10 E R I ECRSISCOSNS O
TNE CTHEN
COH LNOOGL Y
OGRY
E VRI E V
WI E✱W M
✱AY# 1022, 2 0 2 0 2 21 20 ,2200 2✱
# 0AY
M 0 E✱
R I ECRSISCOSNS O
TNE CTHEN
COH LNOOGL Y
OGRY
E VRI E V
WI E W 49
11
✱ THE FUTURE OF CLOUD COMPUTING THE FUTURE OF CLOUD COMPUTING ✱

advanced cloud and AI distributed/edge cloud, many of the major technology acceleration and cloud

the authors
technologies, including the zero-touch automation for vendors, primarily as part of intelligence, taking advantage
use of emerging HW to build orchestration and NFV. the Multimedia, Operations of advanced hardware
disruptive cloud compute Szabo is the coauthor of Support Systems/Business offload capabilities and AI
solutions. Sargor holds a B. more than 80 publications Support Systems and Digital technologies. He holds a B.
Tech. in electrical and holds a both a Ph.D. Services parts of the Ericsson Eng. from Sherbrooke
in electrical engineering and organization, gathering University in Canada.
an MBA from the Budapest insights into many aspects
University of Technology of technology evolution.
Wolfgang John and Economics, Hungary. Chakri Padala Drake currently serves as a
◆ is a research leader and ◆ joined Ericsson in 2007 technology expert in the area
scientist at Ericsson and currently serves as a of platform technologies,
Research in Stockholm. master researcher within working with both technology
His current research focuses Cloud Systems and Platforms intelligence/scouting as well
primarily on distributed cloud at Ericsson Research in as with architecture. He holds
computing systems and engineering from the Indian Bangalore, India. His research a B.Sc. in computer science
platform concepts for both Institute of Technology interests include new from Umeå University in
telco and IT applications. in Mumbai and an M.S. memory/storage Sweden. Miljenko Opsenica
Since joining Ericsson in in computer engineering technologies, acceleration ◆ joined Ericsson in 1998
2011, he has also done from North Carolina State of SW functions and and currently serves as a
research on NFV, software- University in the US. He has Ahsan Javed Awan operating system stacks. master researcher at
defined networking and coauthored more than ◆ is an experienced Padala has an M.S. from the Ericsson Research in
network management. John 25 patents and several researcher of warehouse- University of Louisiana at Finland (NomadicLab),
holds a Ph.D. in computer publications. scale computers at Ericsson Lafayette in the US, and a where he is working on
engineering from Chalmers Research. Prior to joining B.Tech. from the National cloud architectures and
University of Technology in Ericsson in 2018, he worked Institute of Technology in technologies, orchestration
Gothenburg, Sweden, and at Imperial College London, Warangal, India. frameworks and automation.
has coauthored more than IBM Research – Tokyo Opsenica also leads Ericsson
The authors would
like to thank 50 scientific papers and in Japan, and Barcelona Martin Julien Research’s integrated
Jörg Niemöller, reports, as well as several Supercomputing Center ◆ joined Ericsson in 1995 connectivity and edge
Fetahi Wuhib, patent families. in Spain. Awan holds an and currently serves as a program, which focuses on
Andrew Williams, Erasmus Mundus joint Ph.D. senior specialist in cloud integrated edge architecture,
Torbjörn Keisu, Chandramouli from KTH Royal Institute of systems and platforms. cross-resource domain
Daniel Seiler, Sargor Technology in Stockholm, With deep expertise in interactions and performance
Jonas Falkenå, ◆ currently heads the Robert Szabo Sweden and the Universitat distributed systems, optimization management.
Jonas Bjurel, AI-inspired Design team ◆ joined Ericsson in 2013 Politècnica de Catalunya networking and optical He holds an M.Sc. in electrical
Tobias Lindqvist within the Ericsson Global and currently serves as a in Barcelona, Spain, for his interconnects, he has played engineering and computing
and Azimeh AI Accelerator in Bangalore, principal researcher in Cloud work on performance Edvard Drake a significant role in the from the University of Zagreb
Sefidcon for their India. He joined Ericsson in Systems and Platforms at characterization and ◆ joined Ericsson in 1993. development of innovative in Croatia.
contributions to 2007 and previously headed Ericsson Research in optimization of in-memory Since the early 2000s, he has cloud system infrastructure
this article. the Ericsson Research team Hungary. At present, his work data analytics on a scale-up been deeply engaged in products. Julien’s current
in Bangalore with a focus on focuses primarily on server. technology relations with work mainly focuses on cloud

50
12 E R I ECRSISCOSNS O
TNE CTHEN
COH LNOOGL Y
OGRY
E VRI E V
WI E✱W M
✱AY# 1022, 2 0 2 0 2 21 20 ,2200 2✱
# 0AY
M 0 E✱
R I ECRSISCOSNS O
TNE CTHEN
COH LNOOGL Y
OGRY
E VRI E V
WI E W 51
13
✱ CRITICAL IOT CONNECTIVITY CRITICAL IOT CONNECTIVITY ✱

Critical IoT Time-critical use cases

The majority of time-critical use cases can be
Local-area deployment includes both indoor
and outdoor coverage for a small geographical

connectivity
classified into the following four use case families: area such as a port, farm, factory, mine or hospital.
❭ Industrial control Confined wide-area deployment is for a predefined
geographical area – along a highway, between
❭ Mobility automation
certain electrical substations, or within a city
❭ Remote control
center, for example. General wide-area deployment
❭ Real-time media
is about serving devices virtually anywhere.
Each family is relevant for multiple industries and Common to all time-critical use cases is the fact
includes a wide range of use cases with more or that the communication service requirements
IDEAL FOR TIME-CRITICAL COMMUNICATIONS less stringent time-critical requirements, as shown depend on the dynamics of the use case and the
in Figure 1. application implementation. A highly dynamic
Furthermore, there are three main network system requires faster control with shorter round-
deployment scenarios depending on the coverage trip times (RTTs), while a slower control loop is
needs of time-critical services in different industries: sufficient for a system that operates more slowly.
Various factors – such as device processing
Critical Internet of Things (IoT) connectivity is ideal for a wide range ❭ Local area
capabilities, the processing split between the device
❭ Confined wide area
of time-critical use cases across most industry verticals, and mobile ❭ General wide area
and the application server, the application’s ability to
extrapolate and predict data in case of missing
network operators are uniquely positioned to deliver it.

Time-critical use cases common across multiple industries Deployment scenarios

Machine vision Local area

Industrial control Control to control
FREDRIK ALRIKSSON,
Cellular Internet of Things (IoT) is driving with more advanced 5G networks. Industrial in production line for robotics Closed-loop Confined wide area
Open or closed-loop process control General wide area
LISA BOSTRÖM, transformation across various sectors by Automation IoT, the fourth segment, includes control of industrial
Process
PLC to robot controller
automation systems
JOACHIM SACHS, enabling innovative services for consumers capabilities on top of Critical IoT that enable monitoring Smart grid control
Motion control
Industries
Y.-P. ERIC WANG, and enterprises. There are currently more integration of the 5G system with real-time
Mobility automation Automated container Cooperative maneuvering Cloud motion Entertainment
ALI ZAIDI than one billion cellular IoT connections, and Ethernet and time-sensitive networking (TSN) transport in port of vehicles control of AGVs Automotive
Automated control loops
Ericsson forecasts that there will be around used in wired industrial automation networks. for mobile vehicles and Cooperative AGVs in Machine vision for Collaborative Transportation
robots a production line intersection safety mobile robots Health care
five billion connections by 2025 [1]. Critical IoT addresses the time-critical Education
communication needs of individuals, Remote control Media production
Remote control with Remote control with Remote control with Forestry
■ As 5G deployments gain momentum globally, enterprises and public institutions. It is intended Human control of
video/audio AR overlay haptic feedback Public safety
remote devices
enterprises in almost every industry are exploring for time-critical applications that demand data Utilities
Oil & gas
the potential of 5G to transform their products, delivery within a specified time duration with Real-time media Premium experience Cloud-rendered Railways
Cloud-assisted
services and businesses. Since the requirements required guarantee (reliability) levels, such as Real, virtual and basic AR
cloud-assisted AR Interactive VR
cloud gaming
AR Agriculture
Manufacturing
for wireless connectivity in different industries data delivery within 50ms with 99.9 percent combined
environments Cloud gaming Warehousing
Media production
vary, it is useful to group them into four distinct likelihood (reliability). Mining
Ports
IoT connectivity segments: Massive IoT, Critical IoT is a paradigm shift from the enhanced 10s of ms latency 1s of ms latency Construction
Time-criticality
Broadband IoT, Critical IoT and Industrial mobile broadband (eMBB) connectivity, where the 99% reliability 99.999% reliability

Automation IoT [2]. data rate is maximized without any guarantee on

While Massive IoT and Broadband IoT already latency [3]. Many industry sectors have already
exist in 4G networks, Critical IoT will be introduced started piloting time-critical use cases. Figure 1 Examples of use cases enabled by Critical IoT

52
2 E R I ECRSISCOSNS OT N
E CTHE N
COH LNO
OGL Y
OGRY
E VRI E V
WI E✱W J✱U N#E 022, 2 0 2 0 J#U0N2 E2 20,2200 2✱
0 E✱
R I ECRSISCOSN
S OT N
E CTHE N
COH LNO
OGLY
OGRY
E VRI E V
WI E W 53
3
✱ CRITICAL IOT CONNECTIVITY CRITICAL IOT CONNECTIVITY ✱

The communication service requirements for Key network technologies and architectures
remote control depend on how fast the remote Achievable end-to-end (E2E) latencies depend on
Time-critical use case trials environment changes, the required precision of the the available network and compute infrastructure,
In partnership with leading industry partners and mobile network operators, Ericsson has trialed various task and the required QoE. Control-loop latency and software features, and how the use case is imple-
Critical IoT use cases including: audio/video quality are important factors for QoE and mented. In remote control, the physical distance
the ergonomics for the remote operator. Haptic feed- between the remote operator and the teleoperated
❭ Industrial control for manufacturing vehicles: https://www.ericsson.com/en/networks/cases/accelerate- back and augmented reality (AR) can be used to equipment is a physical property of the use case.
factory-automation-with-5g-urllc further improve the operator QoE and task precision, In other use cases, the physical distance between
❭ Industrial control for manufacturing jet engines: https://www.youtube.com/watch?v=XZWC_ttighM and will make the acceptable latencies even stricter. end nodes can be reduced by distributed cloud
❭ Remote control in mining: https://www.youtube.com/watch?v=C4l0UKZ-FCc&t=7s Real-time media comprises use cases where processing, as in AR cloud gaming, where the AR
❭ Remote control of autonomous trucks: https://www.ericsson.com/en/press-releases/2018/11/ericsson-
media is produced and consumed in real time, overlay can be rendered in an edge cloud to limit
einride-and-telia-power-sustainable-self-driving-trucks-with-5g
and delays have a negative impact on QoE. interaction latencies. Network orchestration
Mobile applications for gaming and entertainment, optimizes the placement of network and application
❭ Remote bus driving: https://www.youtube.com/watch?v=lPyzGTD5FtM
including AR and virtual reality (VR), are common, functions to ensure efficient use of the compute
❭ Cooperative vehicle maneuvers: https://5gcar.eu/ with processing and rendering done locally in the and network infrastructure while restricting the
❭ Virtual reality and real-time media: https://www.ericsson.com/en/blog/2017/5/its-all-green-flags-for-5g-at- device. Time-critical communication will make it transmission paths according to latency needs [7].
the-indianapolis-motor-speedway possible to offload parts of the processing and The 5G network comprises two functional domains:
❭ Augmented reality: https://www.ericsson.com/en/news/2018/3/5g-augmented-reality rendering to the cloud [6], thereby improving the the next generation (5G) RAN (NG-RAN) and the
user experience and enabling the use of more 5G Core (5GC), which are built on an underlying
❭ Smart harbor: https://www.ericsson.com/en/press-releases/2019/2/ericsson-and-china-unicom-announce-
lightweight devices (head-mounted, for example). transport network. All three – the NG-RAN, the
5g-smart-harbor-at-the-port-of-qingdao
Time-critical communication can enable cloud 5GC and the transport network – contribute to the
gaming over cellular networks as well as new E2E reliability and latency, which is further affected
applications in sectors such as manufacturing, by the device implementation.
packets, rate adaptivity and which codecs are used in a warehouse and collaborative maneuvering on education, health care and public safety. It is The NG-RAN is deployed in a distributed fashion
– impact both the application RTT and the latency public roads are examples of more time-critical expected to drive more widespread use of mobile to provide radio coverage with good performance,
requirements on the communication network. mobility automation use cases, while the AR and VR. Advanced media production (such as availability and capacity. The 5GC provides
Industrial control includes a very broad set of collaborative mobile robots used in flexible real-time production of live performances) with its connectivity of the device to the external services
applications, present in most industry verticals [4]. production cells represent an even higher degree strict delay and synchronization requirements, and applications. The network latency between the
These applications typically consider late messages of time-criticality. is another area where time-critical communication application and the RAN can be a major contributor
as lost. Process monitoring, controller-to-controller Remote control refers to the remote control can enable new use cases. to E2E latency.
communication between production cells and some of equipment by humans. The ability to remotely
control functions for the electricity grid are examples control equipment is an important step in the
of use cases with modest time-criticality, while use evolution toward autonomous vehicles (to take
cases such as closed-loop process control and temporary control of a driverless bus in scenarios Terms and abbreviations
motion control have very stringent requirements. not covered by its own automation functions) 5GC – 5G Core | AAS – Advanced Antenna System | AGV – Automated Guided Vehicle |
Mobility automation refers to the automation of and for flying drones beyond visual line-of-sight. AR – Augmented Reality | CA – Carrier Aggregation | DC – Data Center | DL – Downlink |
control loops for mobile vehicles and robots. Remote control can also improve work E2E – End-to-End | eMBB – Enhanced Mobile Broadband | FDD – Frequency Division Duplex|
Examples of the least time-critical use cases in this environments and productivity by moving humans IOT – Internet of Things | MNO – Mobile Network Operator | NG-RAN – Next Generation RAN |
category include the relatively self-sufficient out of inconvenient or hazardous environments NPN – Non-Public Network | NR – New Radio | PLC – Programmable Logic Controller | RTT – Round-Trip
automated guided vehicles (AGVs) equipped with – remote-controlled mining equipment [5] Time | TDD – Time Division Duplex | TSN – Time-Sensitive Networking | UE – User Equipment |
advanced on-board sensors that are used for is one example. Such solutions also offer the UL – Uplink | URLLC – Ultra-Reliable Low-Latency Communication | VR – Virtual Reality
transportation in ports and mines. Infrastructure- benefit of providing enterprises with access
assisted vehicles such as fast-moving AGVs to a broader workforce.

54
4 E R I ECRSISCOSNS OT N
E CTHE N
COH LNO
OGL Y
OGRY
E VRI E V
WI E✱W J✱U N#E 022, 2 0 2 0 J#U0N2 E2 20,2200 2✱
0 E✱
R I ECRSISCOSN
S OT N
E CTHE N
COH LNO
OGLY
OGRY
E VRI E V
WI E W 55
5
✱ CRITICAL IOT CONNECTIVITY CRITICAL IOT CONNECTIVITY ✱

NR standard release, the target has been to enable with a suitable QoS flow profile is established,
one-way latencies through the RAN of down to 1ms, according to a corresponding service subscription.
On-premises Edge sites Regional DC National DC where a timely data delivery can be ensured with Larger customers, like an enterprise, are typically
~0-1ms RTT ~5-20ms RTT ~10-40ms RTT
1-5ms RTT
99.999 percent probability. interested in connectivity for an entire device group.
General wide area Features addressing low latency include ultra- For this purpose, 5G has defined non-public networks
short transmissions, instant transmission (NPNs), which are real or virtual networks that are
mechanisms to minimize the waiting time for restricted for usage by an authorized group of
uplink (UL) data, rapid retransmission protocols devices for their private communication [10].
Confined wide area that minimize feedback delays from a receiver An NPN can be realized as a standalone network
to the transmitter, instant preemption and not coupled to a public network that is purpose-
prioritization mechanisms, interruption-free built to provide customer services at the
Local area mobility and fast processing capabilities of customer premises.
devices and base stations. Alternatively, an NPN may share parts of the
Features addressing high reliability include a range network infrastructure with a public network,
of robust signal transmission formats. There are like a common RAN that is shared for private and
Core user plane Subscription data management
methods for duplicate transmissions to improve public users. Beyond the shared RAN, the NPN
National DC
Edge DC Core control plane Application server Alternative options reliability through diversity, both within a carrier may have a separate dedicated core network and
Network exposure Redundant connection (optional)
using transmissions through multiple antenna local breakout – that is, it may be located on the
points, as well as between carriers through either customer’s premises with its own device
carrier aggregation (CA) or multi-connectivity. authentication, service handling and traffic
Figure 2 Examples of network architectures for low latency and/or high reliability
Advanced antenna systems (AAS) have management. Finally, an NPN can be a network
tremendous potential to improve the link budget service that is provided by a mobile network operator
Figure 2 provides examples of network telecommunication functions, the distributed and reduce interference. The vendor-specific radio (MNO) as a customer-specific network slice.
architectures for low latency and/or high reliability, computing platform allows the hosting of network configuration, algorithms for scheduling, Some NPNs may be customized to provide
and illustrates the effect of moving the application application functions in the network [8]. link adaptation, admission and load control that are dedicated functionality for industrial automation,
closer to the device. If an application is hosted in a Network slicing makes it possible to create at the heart of NR make it possible to fulfill service including 5G-LAN services and Ethernet support,
central national data center (DC), the transport multiple logical networks that share a common requirements while ensuring an optimized providing ultra-low deterministic latency,
network round-trip latency can be in the order of network infrastructure. A dedicated network slice utilization of available resources. interworking with IEEE (the Institute of Electrical
10-40ms, depending on the distance to the DC can be created by configuring and connecting Support for highly reliable communication and Electronics Engineers) TSN, and time-
and how well the transport network is built out. computing and networking resources across the has also been addressed for the 5GC, by introducing synchronization to synchronize devices over 5G
Transport latency can be reduced to 5-20ms by radio, transport and core networks. By reserving options for redundant data transmission. to a reference time [11, 12]. Enhanced service
moving applications to a regional DC or even to resources, a high availability of time-critical Multiple redundant user-plane connections exposure of the 5G system makes it possible to better
1-5ms for edge sites. For local network deployments services can be ensured and latencies for with disjoint routes and nodes can be established integrate 5G into an industrial system [13] by means
with networking functions and applications hosted queuing can be avoided. simultaneously. This may include the usage of of service interfaces for device management
on-premises, transport latencies become negligible. Network orchestration automates the creation, separate user equipment (UE) on the different (device onboarding, connectivity management and
Control of the network topology and the transport modification and deletion of slices according to a routes. 5G provides QoS, and by configuring a monitoring, for example) and network management.
latency can be achieved by placing virtualized core slice service requirement [2]. This can imply that suitable QoS flow through the 5G system for
network functions for execution at any location
within the distributed computing platform of the
compute locations are selected according to
guaranteed resource availability and transport
transporting time-critical communication,
queuing latencies due to conflicting traffic can
... [AAS] HAVE TREMENDOUS
network. This software-based design provides latency rather than the lowest compute costs, for be avoided by traffic separation with resource POTENTIAL TO IMPROVE THE
flexibility in updating the network with new reservations and/or traffic prioritization.
functionality and reconfiguring it according to
example. 5G New Radio (NR) provides several
capabilities for ultra-reliable low-latency For a time-critical communication service
LINK BUDGET AND REDUCE
requirements. In addition to running communication (URLLC) [7, 9]. From the first that is requested by a consumer, a data session INTERFERENCE

56
6 E R I ECRSISCOSNS OT N
E CTHE N
COH LNO
OGL Y
OGRY
E VRI E V
WI E✱W J✱U N#E 022, 2 0 2 0 J#U0N2 E2 20,2200 2✱
0 E✱
R I ECRSISCOSN
S OT N
E CTHE N
COH LNO
OGLY
OGRY
E VRI E V
WI E W 57
7
✱ CRITICAL IOT CONNECTIVITY CRITICAL IOT CONNECTIVITY ✱

to a higher consumption of radio resources, as the to have 95 percent indoor UL coverage even using
Spectrum option Frequency allocation Deployment scenario Subcarrier spacing
scheduler needs to provision a larger link adaptation the mid-band carriers, although the achievable
margin to reduce the likelihood of failures in the capacity is limited. Local indoor deployments
Low-band FDD 2x10MHz @ 800MHz Wide area 15kHz
initial transmissions. Furthermore, we observe that are a prerequisite in high-loss or very large buildings,
the mid-band options can offer a significant capacity and are also necessary in other buildings if high
Mid-band FDD 2x20MHz @ 2GHz Wide area 15kHz boost for the wide-area scenario, thanks to large indoor coverage and capacity is desired.
available bandwidths and use of AAS. Although suburban and rural scenarios typically
Mid-band TDD 50MHz @ 3.5GHz Wide area 30kHz Among the two mid-band options studied, FDD have larger cells, it is nonetheless possible to achieve
at 2GHz is attractive when greater UL coverage similar results there. This is because antennas in
Mid-band TDD 100MHz @ 3.5GHz Local factory 30kHz (99 percent) is desired. Our case studies also show suburban and rural environments tend to be
that it is challenging for the wide-area deployment installed at a greater height, there are fewer
High-band TDD 400MHz @ 30GHz Local factory 120kHz to provide full indoor UL Critical IoT coverage obstacles and the smaller buildings result in less
using mid-band spectrum options, due to building- wall-penetration loss. These factors compensate
Table 1 Spectrum assets considered in the case studies penetration loss. In general, indoor coverage for the differences in cell range, making it feasible
depends on building materials and building sizes. to achieve very good Critical IoT performance
Under favorable conditions, such as low-loss in suburban and rural scenarios as well.
5G spectrum flexibility Case studies facades and limited building sizes (that is, For local-area studies (scenario #2), the
5G NR allows MNOs to take full advantage To illustrate how 5G spectrum assets can be less than 3,600sq m in footprint), it is feasible deployment using 3.5GHz spectrum is based
of all available spectrum assets. NR can be utilized for Critical IoT, we have put together
deployed using the spectrum assets used for case studies for two deployment scenarios:
the LTE networks, either through refarming wide-area deployment and local-area Scenario #1: Central London wide-area
or spectrum sharing [14]. Most of the LTE deployment inside a factory.
spectrum assets are in the low and mid bands, The wide-area scenario is based on a macro- Downlink traffic [Mbps] Uplink traffic [Mbps]
4646 4545
118 117
which in the 5G era will continue to be used deployment in central London with an inter-site 140
120
112
87
101
85 92
50
45
40
36 35
35 29
73 83 22
for wide-area coverage. Traffic growth will distance of approximately 450m, assuming low- 100
80 64 59 63
41 54
55 61
42
35
30
25
29 24
18
312717 30
26 15

drive the need for increased network capacity band FDD, mid-band FDD and mid-band TDD 60
40
39 37 30
31
26
27
21
20
15
1715 15
6 13 1
10 433 332 2 2 1 1 1 0.2
throughout the 5G era. spectrum options. For the mid-band deployments, 20 4 4 4 4 3 2 2 2 2 1
0
2 0.2 5
0

Increased capacity can be achieved by adding we include an AAS, with eight antenna columns for 99% 99.9% 99% 99.9% 99% 99.9% 99% 99.9% 99% 99.9% 99% 99.9% 99% 99.9% 99% 99.9% 99% 99.9% 99% 99.9% 99% 99.9% 99% 99.9%
24ms 8ms 24ms 8ms 24ms 8ms 24ms 8ms 24ms 8ms 24ms 8ms
more spectrum assets, densifying the network and/ 3.5GHz and four for 2GHz. Devices with four 800MHz 2GHz 3.5GHz 800MHz 2GHz 3.5GHz

or upgrading capabilities at existing sites. New 5G receiver branches are used in the evaluation. 90% coverage 95% coverage 99% coverage 90% coverage 95% coverage 99% coverage

spectrum options in the mid bands (around 3.5GHz) The local factory setup is based on a factory Scenario #2 Factory indoor deployment
and in the high bands (such as the millimeter wave automation scenario [15] and assumes mid-band
frequencies) present great opportunities with large Downlink traffic [Mbps] Uplink traffic [Mbps]
and high-band TDD options. Table 1 lists the 450 450 395
bandwidths. spectrum options chosen in the case studies. 400
350
389
316
400
350 308
300 300
Operating with these new spectrum assets, the The top half of Figure 3 presents the served 250
200 189
250
200 160 160 160 165
added RAN nodes can also use advanced hardware capacity per cell versus various reliability and 150
100
140 140 140 140 125 150
100 80 114
50 50
features such as an AAS to fully capitalize on the round-trip RAN latency requirements for outdoor 0 0
99.9% 99.999% 99.9% 99.999% 99.9% 99.999% 99.9% 99.999% 99.9% 99.999% 99.9% 99.999% 99.9% 99.999% 99.9% 99.999%
benefits of NR. The coverage provided by the UEs in the central London wide-area deployment 5ms 2ms 5ms 2ms 5ms 2ms 5ms 2ms
low-band and mid-band spectrum assets is key to scenario. All the TDD cases assume a TDD pattern 3.5GHz 30GHz 3.5GHz 30GHz

enable Critical IoT services in wide-area deployments. with 3:1 downlink (DL) and UL split. Observe the 100% coverage 100% coverage

Adding network capacity over time will not only cost in terms of capacity when pushing for tighter
increase the capacity for eMBB, but also boost reliability and latency requirements. Generally,
the capacity for Critical IoT. a tighter reliability or latency requirement leads Figure 3 Served capacity per cell versus various reliability and round-trip RAN latency requirements for the two scenarios

58
8 E R I ECRSISCOSNS OT N
E CTHE N
COH LNO
OGL Y
OGRY
E VRI E V
WI E✱W J✱U N#E 022, 2 0 2 0 J#U0N2 E2 20,2200 2✱
0 E✱
R I ECRSISCOSN
S OT N
E CTHE N
COH LNO
OGLY
OGRY
E VRI E V
WI E W 59
9
✱ CRITICAL IOT CONNECTIVITY CRITICAL IOT CONNECTIVITY ✱

better carrier, depending on the operating condition, In some regions, significant TDD spectrum has Conclusion
MNOs CAN START user location and use case requirements. been allocated to enterprises for local use (in the Critical Internet of Things connectivity addresses
TO ADDRESS TIME-CRITICAL With a low-band carrier, there are also benefits order of 100MHz) in mid-band and millimeter-wave time-critical communication needs across
of pooling an FDD carrier and a TDD carrier from frequency ranges. For both confined wide-area and various industries, enabling innovative services
USE CASES ... THROUGH a latency point of view, using the FDD carrier local-area scenarios, the reuse of MNOs’ existing for consumers and enterprises. Mobile network
SOFTWARE UPGRADES to mitigate the extra alignment delay introduced infrastructure and their flexible spectrum assets operators are uniquely positioned to enable
on a TDD carrier due to the DL-UL pattern. (in combination with dedicated spectrum, if available) time-critical services with advanced 5G networks
on a single-cell deployment with eight antennas brings major value and opportunities. This approach in a systematic and cost-effective way, taking
installed in the ceiling, uniformly distributed across Deployment strategy makes it possible to exploit the full potential of various full advantage of flexible spectrum assets,
the entire factory, and a DL and UL symmetric MNOs have started to upgrade some 4G LTE band combinations and support seamless mobility efficient reuse of existing footprint and flexible
TDD pattern. For the high-band deployment, radio base station equipment to 5G NR through and interaction between public and dedicated software-based network design.
eight transmission points with full frequency reuse software upgrades. The dynamic spectrum communication infrastructure.
are considered. The 3GPP indoor factory channel sharing solution allows efficient coexistence
model with dense clusters, including machinery, of LTE and NR in the same spectrum band
assembly lines, storage shelves and so on [16], is used. down to millisecond level [14].
To achieve 2ms round-trip RAN latency, NR mini- MNOs can start to address time-critical use cases
slot and configured grant features are used. (Using in the wide area (the entertainment, health care and Further reading
the same features, an FDD carrier with 15kHz education sectors, for example) by adding support ❭ Ericsson, Evolving Cellular IoT for industry digitalization, available at: https://www.ericsson.com/en/
subcarrier spacing can also achieve similar latency.) for Critical IoT connectivity to the NR carriers networks/offerings/cellular-iot
The bottom half of Figure 3 shows that both through software upgrades. More stringent, time-
❭ Ericsson, IoT connectivity, available at: https://www.ericsson.com/en/internet-of-things/iot-connectivity
DL and UL traffic achieve 100 percent coverage. critical requirements call for radio network
Tightening Critical IoT requirements reduces densification, edge computing, and further
capacity, however, and this is more evident in the distribution and duplication of core network
high-band case. For the mid-band case, all users functions, which can be done gradually over time,
consistently reach the highest spectral efficiency while maximizing returns on investment.
except for UL traffic with the most stringent In the confined wide-area scenarios (railways,
requirements, due to good coverage and the utilities, public transport and the like), relatively
absence of interference achieved by the single-cell stringent requirements can be addressed with
distributed antenna deployment. reasonable investments in existing and new
5G NR CA allows radio resources from multiple infrastructure. In local-area scenarios such as
carriers in multiple bands to be pooled to serve a factories, ports and mines, even extreme time-
user. For example, DL traffic can be delivered using critical requirements can be supported once the
a mid-band carrier even when the UL service E2E ecosystem is established.
requirements are not attainable on that mid-band Dedicated spectrum has been allocated to some
carrier, by using a low-band carrier for the UL industry sectors in certain regions. In the wide-area
control and data traffic. This allows the DL capacity of scenarios such as public safety and railways, the
the mid-band carrier to be utilized to a greater extent. allocated bandwidths are typically small (10MHz
In essence, inter-band CA allows an MNO to or below) and unable to meet the capacity demands
improve coverage, spectral efficiency and capacity of emerging use cases, especially those with time-
by dynamically directing the traffic through the critical requirements.

60
10 E R I ECRSISCOSNS OT N
E CTHE N
COH LNO
OGL Y
OGRY
E VRI E V
WI E✱W J✱U N#E 022, 2 0 2 0 J#U0N2 E2 20,2200 2✱
0 E✱
R I ECRSISCOSN
S OT N
E CTHE N
COH LNO
OGLY
OGRY
E VRI E V
WI E W 61
11
✱ CRITICAL IOT CONNECTIVITY CRITICAL IOT CONNECTIVITY ✱

ization. Boström holds an M. the University of Michigan

References

the authOrs
Sc. in media engineering (Ann Arbor) in the US. Wang
1. Ericsson Mobility Report, November 2019, available at: https://www.ericsson.com/en/mobility-report/ from Luleå University of is coauthor of the book
reports/november-2019 Technology in Sweden. Cellular Internet of Things:
2. Ericsson white paper, Cellular IoT in the 5G era, February 2020, available at: https://www.ericsson.com/en/ From Massive Deployments
reports-and-papers/white-papers/cellular-iot-in-the-5g-era to Critical 5G Applications.
3. 3GPP TR38.913, Study on Scenarios and Requirements for Next Generation Access Technologies, 2017,
available at: http://www.3gpp.org/ftp//Specs/archive/38_series/38.913/38913-e30.zip
4. 5G-ACIA white paper, 5G for Automation in Industry – Primary use cases, functions and service
Fredrik Alriksson
requirements, July 2019, available at: https://www.5g-acia.org/fileadmin/5G-ACIA/Publikationen/5G-ACIA_
◆ is a researcher at
White_Paper_5G_for_Automation_in_Industry/WP_5G_for_Automation_in_Industry_final.pdf
Development Unit Networks,
5. Ericsson Consumer and IndustryLab Insight Report, A case study on automation in mining, June 2018, where he leads strategic
available at: https://www.ericsson.com/en/reports-and-papers/consumerlab/reports/a-case-study-on- technology and concept
automation-in-mining
development within IoT & Joachim Sachs
6. GSMA, Cloud AR/VR Whitepaper, May 8, 2019, available at: https://www.gsma.com/futurenetworks/ New Industries. He joined ◆ is a principal researcher at
resources/gsma-online-document-cloud-ar-vr-whitepaper/ Ericsson in 1999 and has Ericsson Research in Ali Zaidi
7. Proceedings of the IEEE, vol. 107, Issue 2, pp. 325-349, Adaptive 5G Low-Latency Communication for worked in R&D with Stockholm and coordinates ◆ is a strategic product
Tactile Internet Services, February 2019, Sachs, J. et al., available at: http://ieeexplore.ieee.org/stamp/ architecture evolution research activities on 5G for manager for Cellular IoT at
stamp.jsp?tp=&arnumber=8454733&isnumber=8626773 covering a broad set of industrial IoT solutions and Ericsson and also serves as
8. Ericsson Technology Review, Distributed cloud – a key enabler of automotive and industry 4.0 use cases, technology areas including cross-industry research the company’s head of IoT
November 20, 2018, Boberg, C; Svensson, M; Kovács, B, available at: https://www.ericsson.com/en/reports- RAN, Core, IMS and VoLTE. collaborations. He holds a Competence. He holds a
and-papers/ericsson-technology-review/articles/distributed-cloud Alriksson holds an M.Sc. in Ph.D. from the Technical Ph.D. in telecommunications
electrical engineering from University of Berlin in from KTH Royal Institute of
9. Academic Press, Cellular Internet of Things – From Massive Deployments to Critical 5G Applications,
KTH Royal Institute of Germany. Sachs is coauthor Technology in Stockholm.
October 2019, Liberg, O; Sundberg, M; Wang, E; Bergman, J; Sachs, J; Wikström, G, available at:
Technology in Stockholm, of the book Cellular Internet Since joining Ericsson in
https://www.elsevier.com/books/cellular-internet-of-things/liberg/978-0-08-102902-2
Sweden of Things: From Massive 2014, he has been working
10. NGMN white paper, 5G E2E Technology to Support Verticals' URLLC Requirements, November 18, 2019, The authors would
Deployments to Critical 5G with technology and business
available at: https://www.ngmn.org/publications/5g-e2e-technology-to-support-verticals-urllc-requirements.html like to thank
Applications. development of 4G and 5G
11. Ericsson Technology Review, Boosting smart manufacturing with 5G wireless connectivity, January 2019, Yanpeng Yang,
radio access at Ericsson.
available at: https://www.ericsson.com/en/reports-and-papers/ericsson-technology-review/articles/boosting- Anders Furuskär,
Zaidi is currently responsible
smart-manufacturing-with-5g-wireless-connectivity Kittipong
for LTE-M, URLLC, Industrial
Kittichokechai,
12. Ericsson Technology Review, 5G-TSN integration meets networking requirements for industrial IoT, vehicle-to-everything
Anders Bränneby,
automation, August 2019, Farkas, J; Varga, B; Miklós, G; Sachs; J, available at: https://www.ericsson.com/ and local industrial networks.
Fedor
en/ericsson-technology-review/archive/2019/5g-tsn-integration-for-industrial-automation
Chernogorov,
13. 5G-ACIA, Exposure of 5G Capabilities for Connected Industries and Automation Applications Gustav Wikström,
(white paper), June 2020, available at: https://www.5g-acia.org/publications/ Jari Vikberg, Lisa Boström
Mattias Andersson, ◆ is a researcher at
14. Ericsson Spectrum Sharing, available at: https://www.ericsson.com/en/networks/offerings/5g/sharing-
Ralf Keller, Development Unit Networks,
spectrum-with-ericsson-spectrum-sharing
Kun Wang, where she does research and Y.-P. Eric Wang
15. 3GPP TR 38.824, Study on physical layer enhancements for NR URLLC, 2019, available at:
Torsten Dudda and concept development within ◆ joined Ericsson in 1995
http://www.3gpp.org/ftp//Specs/archive/38_series/38.824/38824-g00.zip
Marie Hogan for IoT & New Industries. She and is currently a principal
16. 3GPP TR 38.901, Study on channel model for frequencies from 0.5 to 100 GHz, 2019, available at: their contributions joined Ericsson in 2006 and researcher at Ericsson
http://www.3gpp.org/ftp//Specs/archive/38_series/38.901/38901-g00.zip to this article. has worked extensively with Research. He holds a Ph.D. in
RAN R&D and standard- electrical engineering from

62
12 E R I ECRSISCOSNS OT N
E CTHE N
COH LNO
OGL Y
OGRY
E VRI E V
WI E✱W J✱U N#E 022, 2 0 2 0 J#U0N2 E2 20,2200 2✱
0 E✱
R I ECRSISCOSN
S OT N
E CTHE N
COH LNO
OGLY
OGRY
E VRI E V
WI E W 63
13
✱ INTEGRATED ACCESS AND BACKHAUL INTEGRATED ACCESS AND BACKHAUL ✱

Integrated access These assets will be deployed on macro sites

(rooftops, towers) and street sites (poles, walls, strands)
in urban areas with high demands on capacity and
speed, as well as in suburban areas with fiber-like
A solution more like IAB was studied for
LTE in 3GPP release 10 in 2011, also known as
LTE relaying [4], but it never gained any commercial
interest. However, with the wide mmWave

and backhaul
fixed wireless access (FWA) services [3]. IAB could bandwidths now becoming available, there is
provide fast deployment of mmWave backhaul for considerable interest in an IAB solution for 5G NR.
new multiband street sites, with an easy migration The work on IAB has been going on in the 3GPP
to fiber-based backhaul if, and when, needed. since 2017, and it is currently being standardized for
release 16, targeting completion during 2020 [5, 6].
Using radio-access technology IAB can provide flexible and scalable multi-hop
to provide backhaul backhauling, using the same or different frequency
Access spectrum has historically been too valuable bands for access and backhaul, as shown in
A NEW TYPE OF WIRELESS BACKHAUL IN 5G and limited to use for backhauling. Its rare use section B of Figure 1.
today is for LTE solutions that provide a single The backhaul is efficiently forwarded across
backhaul hop using a separate frequency band the wirelessly interconnected radio nodes,
from access, as shown in section A of Figure 1. with the backhaul links terminated by an
This approach uses a fixed wireless terminal (FWT) IAB mobile termination (IAB-MT) function.
to provide connectivity to a separate backhaul The IAB-MT could either use a separate antenna
5G New Radio introduces a new type of wireless backhaul known as core instance. The instance could either be in the or share the access antenna of the base station
core for radio access or distributed closer to the (virtual IAB-MT). The latter provides the
integrated access and backhaul that is of particular interest for dense radio nodes to support lower latency inter-site ultimate level of integration, as well as utilizing
deployment of street-level radio nodes. connectivity. It is also possible to use 5G NR to the high-performance base station antennas
provide such separate access and backhaul solutions. for backhaul over longer distances.

HENRIK RONKAINEN,
The combination of millimeter wave (mmWave) two decades and will remain a highly attractive A) Separate access and backhaul Use case examples
JONAS EDSTAM, spectrum – which is becoming available complement to fiber for 5G transport [2]. f1 f2
Urban
ANDERS ERICSSON, globally for 5G – with other spectrum assets Network densification using street-site FWT
Backhaul
CHRISTER ÖSTBERG below 6GHz results in high speeds and deployments comes with new challenges, however. Core
core instance
capacities. The mmWave radio resources can The allowed space and weight for equipment is
only provide limited coverage, though, which limited. The installation, integration and operation f1 f2
makes it reasonable to expect a fairly low level must be simplified with a high degree of automation Backhaul FWT
Core
core instance
of utilization. As a result, there is an opportunity to achieve cost-efficient deployment of RAN Suburban
to use an innovative type of wireless backhaul and transport. This calls for a new type of
in 5G – integrated access and backhaul (IAB) wireless backhaul that is fully integrated with B) Integrated access and backhaul
– to densify networks with multi-band radio 5G New Radio (NR) access. This is where IAB f1 f1 f1 f2
sites at street level. enters the frame. IAB-MT IAB-MT
Core
Core
More than 10GHz of total bandwidth in the
■ Transport networks play a vital role in RANs mmWave frequency range of 24.25GHz to 71GHz Virtual IAB-MT Virtual IAB-MT Indoor
by connecting all the pieces. The use of dark fiber was globally identified for 5G at the ITU World f1 f2
f1 f1
for 5G transport is of growing importance [1], Radio Conference 2019. Already today, 5GHz
Core
and wireless backhaul is an essential complement of mmWave bandwidth is available in the US.
for sites where fiber is either not available or too The best overall performance at the lowest total
costly. In fact, microwave backhaul has been cost of ownership is achieved by using mmWave
the dominant global backhaul media for over in combination with spectrum assets below 6GHz [1]. Figure 1 Solutions using radio-access technology to provide backhaul

64
2 E R I CESRSI O
CNS STOEN
C HT N
ECOHL O
NOG LYORGEYV IREEWV I✱E WJ U✱N E#2032, 2 0 2 0 J#U0N2 E2 2032,02 0✱
2 0E R✱I CESRSI O
CNS STOEN
C HT N
ECOHL O
NOG LYORGEYV IREEWV I E W 65
3
✱ INTEGRATED ACCESS AND BACKHAUL INTEGRATED ACCESS AND BACKHAUL ✱

All IAB-nodes and donor DU(s) that use the same is therefore required between transmission
IAB donor IAB node 1 IAB node 2 CU are part of one gNB, in accordance with the CU/ and reception phases within each IAB node.
DU split architecture. Hence, the wireless backhaul IAB is expected to be of most benefit in
IAB in the 5G architecture
is isolated inside the gNB, and any internal topology, mmWave spectrum, where TDD [8] is used
5GC CU
F1 -U/C
Donor DU Uu
IAB-MT DU Uu
IAB-MT DU routing or backhaul changes can be made without and operators typically have large bandwidth.
impacting the 5GC or neighboring gNBs. A similar A TDD network is typically configured with a
F1 F1
O&M
Other
BAP BAP BAP BAP O&M
Other Forwarding with backhaul
situation is valid for the UEs, for which the IAB node (often regulated) pattern for the time domain
RLC RLC RLC RLC
adaptation protocol (BAP) appears as a normal base station, supporting both allocation of downlink (DL) and uplink (UL)
gNB MAC MAC MAC MAC
NR standalone and non-standalone mode. resources, and an additional level of pattern
Downstream
Parent Child Parent Child As shown in Figure 2, the NR backhaul link is must be used to support combined access and
Upstream
NR backhaul links and roles between a “parent” on the network side and a “child” backhaul traffic. This is illustrated in the example
at the other end. The DU at the parent schedules the in Figure 2, with five different repeated IAB time
Phase 1 DL blocked backhaul downstream and upstream traffic to/from phases for node-local TDD states, where phases
Phase 2 DL blocked DL blocked the IAB-MT at the child, supporting a limited subset 1-4 are mapped to the DL and phase 5 to the UL.
DL
of the NR UE functionality. This includes lower
TDD phases for in-band IAB
Phase 3 DL blocked with half-duplex constraint protocol layer functionality to the parent as well as IAB IS EXPECTED TO
Radio Resource Control and non-access stratum
Phase 4
functionality to the IAB donor CU and 5GC. BE OF MOST BENEFIT
UL Phase 5 The backhaul adaptation protocol (BAP) [11] IN MMWAVE SPECTRUM
enables efficient IP data forwarding across the IAB
interconnected radio nodes, where the BAP data is The mix and duration of different phases can be
Figure 2 The 3GPP IAB concept carried by backhaul Radio Link Control (RLC) flexible depending on the scenario, access/backhaul
channels on each NR backhaul link. Multiple channels link performance, load and so on. Due to the half-
The IAB concept is defined by the 3GPP to be unit (CU) and one or more distributed units (DU). can be configured to enable traffic prioritization and duplex constraint, there will be time periods in which
flexible and scalable to support other use cases The CU hosts the higher layer protocols to the UE QoS enforcement and, based on these properties, the nodes are blocked from transmission in a normal
beyond the initial market interest, such as low-power and terminates the control plane and user plane the BAP entity in each node maps protocol data DL slot, effectively reducing the peak rate for an IAB
indoor radio nodes. There is also research on future interfaces to the 5GC. The CU controls the units to the appropriate backhaul RLC channel. node compared with a similar node with wired (non-
advanced enhancement and optimizations for more DU nodes over the F1 interface(s), where the Hop-by-hop forwarding, from the IAB donor limited) backhaul. This occurs whenever there is a
visionary IAB use. DU node hosts the lower layers for the NR Uu to the destination IAB node, is based on the BAP transmission over the NR backhaul link, as the
interface to the UE. routing identity set by the IAB donor. Any IP traffic receiving end of the link will not operate according to
The 3GPP concept of integrated access As illustrated in Figure 2, the CU/DU split can be forwarded over the BAP, such as F1 and the overall TDD pattern. In the example in Figure 2,
and backhaul architecture is used for IAB and enables efficient operation and maintenance (O&M) of the IAB the backhaul transmission occurs in phases 1-3, and
IAB is defined to reuse existing 5G NR functions multi-hop support. The architecture eliminates nodes, as well as connectivity of any other equipment the normal DL operation is blocked for the receiving
and interfaces, as well as to minimize impact the backhaul core instance at every IAB node shown at the IAB-node site, as shown in Figure 2. nodes (in all sectors) during these phases.
on the core network. The architecture is scalable, in Figure 1A and related overhead due to tunnels The parent node schedules all traffic over the
so that the number of backhaul hops is only inside tunnels, which would become severe for Physical layer aspects backhaul link (phases 1-3) in the same way as
limited by network performance. From a transport large multi-hop chains. The IAB feature is intended to support out-of- for UE scheduling, where frequency division
perspective, IAB provides generic IP connectivity As the time-critical functionality is located in each band and in-band backhauling, where the latter multiplexing or space division multiplexing can
to enable an easy upgrade to fiber transport DU, the F1 interface is well suited for a non-ideal means usage of the same carrier frequencies for be used to separate simultaneous transmissions.
when needed. backhaul such as IAB. The IAB donor is a logical both the NR backhaul links and the access links.
In the 5G network, the gNB base station provides node that provides the NR-based wireless backhaul In-band operation comes with a half-duplex Deployment constraints for integrated access
NR protocol terminations to the user equipment and consists of a CU and wire-connected donor constraint, implying that the IAB-MT part of an and backhaul
(UE) and is connected to the 5G Core (5GC) DU(s). The IAB nodes, which may serve multiple IAB node cannot receive while its collocated DU From a 3GPP architecture perspective, the IAB
network. As defined in 3GPP TS 38.401 [7], the gNB radio sectors, are wireless backhauled to the IAB is transmitting and vice versa to avoid intra-site feature is flexible, supporting multi-hop and a variety
is a logical node, which may be split into one central donor and consist of a DU and an IAB-MT. interference. A strict time-domain separation of topologies. However, there are other aspects that

66
4 E R I CESRSI O
CNS STOEN
C HT N
ECOHL O
NOG LYORGEYV IREEWV I✱E WJ U✱N E#2032, 2 0 2 0 J#U0N2 E2 2032,02 0✱
2 0E R✱I CESRSI O
CNS STOEN
C HT N
ECOHL O
NOG LYORGEYV IREEWV I E W 67
5
✱ INTEGRATED ACCESS AND BACKHAUL INTEGRATED ACCESS AND BACKHAUL ✱

restrict the size of the IAB network topology, Wide-area and local-area IAB-MT are intended
where in-band operation (sharing spectrum for for different deployment scenarios and use differing
m
both backhaul and access) is an essential reason TDD patterns. In Figure 2, all backhaul link traffic - ~450
Achievable downstream backhaul rate
~100m
for these limitations. Larger IAB topologies might is scheduled during DL timeslots but an alternative 400
100% 6%
also require complex control functions. But since TDD scheme may be applied where the UL timeslots 14%
21%
IAB is a complement to fiber, the size of most are used for upstream backhaul. The latter scheme 15%
IAB networks is expected to be small. is restricted in terms of output power, making it

Percentage of NR backhaul links

In a multi-hop network, the first backhaul hop more suitable for local-area deployments. 14%
Urban scenarios
must carry the backhaul bandwidth not only for the The IAB backhaul links give rise to a semi- ~1-3 IAB nodes per macro sector, multi hop 44%
first IAB node, but also for all other IAB nodes synchronous TDD operation, for which the 55%
further down in the hop chain. Deploying multi-hop regulatory framework for local coordination
m <0.5Gbps
networks will therefore eventually lead to backhaul- between operators is not yet in place in all countries 18
00
7%
limited nodes due to congestion in the first hop. [10]. As illustrated by the TDD phases in Figure 2, 65%
0.5-1Gbps

Increasing the number of hops will also increase during certain time slots the IAB node will 14% 1-1.5Gbps

the end-to-end latency and raise the complexity operate in an inverted mode with respect to the 7% 1.5-2Gbps
for scheduling and routing to satisfy QoS. general TDD pattern. This means that a node 24% 2-2.5Gbps
The 3GPP gNB synchronization requirements may be in receiving mode during a DL slot for 14%
>2.5Gbps
apply also for IAB nodes that may be fulfilled with a backhaul link reception and thus suffer from 0%
node-local synchronization solution based on the neighbor node interference, both within the ~800m Urban Suburban Suburban
Both suburban scenarios 15% foliage 23% foliage
Global Navigation Satellite System. In some situations, same channel as well as between channels One IAB node per macro sector, single hop
this is neither wanted nor feasible. Over-the-air in the same frequency band.
synchronization is therefore an alternative option, Even though the backhaul link is more robust
using periodic parent-transmitted reference symbols against interference due to good link budget, Figure 3 Simulated I AB performance for three scenarios
as the synchronization source for the receiving child measures such as isolation between nodes
node. This scheme implies that the clock accuracy (separation distance, for example) or coordinated
at the donor DU must be better than the 3GPP TDD patterns may still be required to avoid Instead, the backhaul dimensioning for IAB The role of integrated access and backhaul
requirement, as the synchronization budget is shared/ excessive interference. systems needs to be an integrated part of RAN in network evolution
aggregated for all nodes using this donor DU. There are dimensioning, considering the shared radio Densification of current networks will mainly take
therefore several practical reasons to limit the number resources for backhaul and access. From a transport place in urban and dense suburban environments.
of hops and not deploy oversized IAB topologies. THE SIZE OF MOST IAB network perspective, the IAB nodes appear as As one part of assessing the role of IAB, we have
Regardless of topology, there are also general NETWORKS IS EXPECTED extensions of the IAB donor. The same IP performed radio network simulations of such
radio aspects to consider. The 3GPP specifies radio assignment methods can be used for IAB nodes scenarios, as illustrated in Figure 3.
interface requirements for the IAB-MT [8], with two TO BE SMALL as for fiber-connected radio nodes, which facilitate When providing broadband to a home, FWA is a
categories to distinguish different use cases and an easy upgrade to fiber transport when needed. good alternative to fiber in many cases, as it lowers
characteristics. One category is for wide-area usage Integrated access and backhaul IAB can also provide IP connectivity for other the barrier to entry and supports faster deployment
with planned site deployment, such as backhaul of from a backhaul perspective equipment at the IAB node site, as shown in Figure 2. [3]. In cases where the traffic demands require
street sites; the other is for local-area usage with Traditional backhaul is a service provided by the The transport performance requirements densification – in dense suburban areas in the US,
site deployments that may not be preplanned. transport network domain to the radio-access nodes. to the IAB donor are affected by the connected for example – the use of wireless backhaul can
The wide-area category enables an integrated For IAB, a segment of the backhaul is embedded in IAB nodes. The busy hour data traffic is funneled further add to these advantages.
solution for the access and backhaul links, where the the RAN domain, sharing common radio resources. through the IAB donor and increases with each We studied FWA using IAB in simulations of two
IAB node can benefit from using the full base station The backhaul transport cannot be dimensioned connected IAB node. The latency and US suburban neighborhoods in the San Francisco
capabilities – such as advanced antenna systems on an individual node basis, as the IAB donor synchronization requirements for the transport Bay Area, with the level of foliage as the main difference:
(AAS) [9] and high output power – to provide good terminates the “common backhaul” for all underlying are also affected, as each IAB backhaul hop adds 15 and 23 percent respectively. As a reference, our
backhaul link performance and relatively large IAB nodes extending the radio access to UEs latency and timing error. These aspects will also estimates indicate that about half of the dense suburban
distance between parent and child nodes. through a network of backhaul and access links. limit the size of the IAB topology. areas in the US have a foliage level lower than 15 percent.

68
6 E R I CESRSI O
CNS STOEN
C HT N
ECOHL O
NOG LYORGEYV IREEWV I✱E WJ U✱N E#2032, 2 0 2 0 J#U0N2 E2 2032,02 0✱
2 0E R✱I CESRSI O
CNS STOEN
C HT N
ECOHL O
NOG LYORGEYV IREEWV I E W 69
7
✱ INTEGRATED ACCESS AND BACKHAUL INTEGRATED ACCESS AND BACKHAUL ✱

Both areas had a macro grid, with an inter-site access and backhaul radio resources. Fiber backhaul
distance of 1,800m and three sectors per site. In is therefore recommended for such sites. For the PoC-UE
order to serve households with an average data remaining 40 percent of the sites, IAB could be a
consumption of 1,000GB per month, a densification viable option, despite the amount of foliage.
was made where each macro sector provided In a simulation of urban London, a densification PoC-IAB node #1
backhaul to a street site with a single hop of about with street sites is required to extend coverage and
800m, as shown in Figure 3. improve mobile broadband capacity both indoors
All sites were deployed with 40MHz on mid band and outdoors. All sites use mid band and mmWave
for access and 800MHz on mmWave for access and for access, and the mmWave is also used for
backhaul. The locations of the new street sites were backhauling between street sites and macro sites.
chosen to secure good backhaul links from the The backhaul topology was a tree structure with one
macro sites as well as good access coverage to the to four hops, where most sites only had a single hop.
homes. An ideal position can be a utility pole in line- The simulations show that the need for an LOS
of-sight (LOS) of the macro site with a surrounding backhaul link is less critical in urban environments PoC-IAB node #2
area with few or no obstacles. Moreover, the risk of than in suburban ones, thanks to strong reflections
future infrastructure changes blocking the backhaul in the city environment, making it relatively easy to
link should be considered. find locations with good signal strength.
In the area with less foliage, the street sites off- Furthermore, the backhaul links are shorter in an
load the macro sites by serving around 40 percent of urban environment, and the impact of foliage is
PoC-IAB donor node
the households. About 80 percent of the backhaul typically less significant due to fewer trees. Figure 3
links have a downstream rate above 2Gbps, as shows the achievable downstream backhaul link
shown in Figure 3. Over 200 households per square rates for the urban case, which are all above 1Gbps.
Figure 4 Deployment of the IAB testbed in Stockholm
kilometer could be served without IAB causing any Eighty percent are above 2Gbps. The densified
traffic limitation, even during peak hours. network provides excellent coverage and capacity for
In the area with more foliage, the propagation both outdoor and indoor users, even though IAB As the backhaul link uses parts of the available Conclusion
conditions are worse both for access and backhaul. consumes part of the spectrum. spectrum resources, the typical rates for users in cells The massive amount of mmWave spectrum
Therefore, additional street sites may be needed to For both suburban and urban scenarios, these served by donors or IAB nodes will be lower than if that is becoming available globally will spark
meet the required capacity, which will affect the simulations show that IAB is an attractive all nodes are fiber connected. Still, with the small a wide variety of innovative 5G use cases.
business case. It was more challenging to find street complement to fiber, with the ability to provide scale network topologies used in the simulations, Integrated access and backhaul (IAB)
site locations with good access as well as backhaul. backhaul in the early years until traffic growth we achieve peak user throughputs far above 1Gbps. is one such innovation that could enhance
Around 60 percent of the backhaul links have a requires all radio resources to be used for access. 5G New Radio to support not only access
downstream rate below 1Gbps, which means the Depending on the subscriber distribution, IAB may Proof of concept but also wireless backhaul.
backhaul will consume a large part of the common not even need to be replaced by fiber at some sites. In order to properly assess IAB-specific Our radio network simulations show that
performance aspects, Ericsson has developed IAB could serve as a versatile backhaul option
an IAB proof of concept (PoC) testbed in an for street sites in urban and suburban areas, using
authentic environment. At the Ericsson site in small-scale star and tree backhaul topologies.
Stockholm we have set up a two-hop IAB It could also be useful for temporary deployments
Terms and abbreviations deployment using two-sector IAB nodes with for special events or emergency situations.
5GC – 5G Core | AAS – Advanced Antenna System | BAP – Backhaul Adaptation Protocol | either 28GHz or 39GHz AAS radios with Point-to-point microwave backhaul will remain
CU – Central Unit | DL – Downlink | DU – Distributed Unit | F1 – Interface CU–DU | FWA – Fixed Wireless 100MHz bandwidth, as shown in Figure 4. an essential complement to fiber for 5G transport
Access | FWT – Fixed Wireless Terminal | gNB – gNodeB | IAB – Integrated Access and Backhaul | The IAB nodes are in the wide-area category with for traditional macro sites, while IAB is a promising
LOS – Line-of-Sight | MAC – Medium Access Control | mmWave – Millimeter Wave | MT – Mobile Termination |
NR backhaul links using DL slots for all transmitted advanced concept that may become as important
NR – New Radio | O&M – Operation and Maintenance | PoC – Proof of Concept | RLC – Radio Link Control |
data. Initial test results are aligned with expectations for wireless backhaul of street sites.
TDD – Time Division Duplex | UE – User Equipment | UL – Uplink | Uu – radio interface between RAN and UE
and show backhaul bitrates near the predicted
and end-to-end peak rates independent of hop level.

70
8 E R I CESRSI O
CNS STOEN
C HT N
ECOHL O
NOG LYORGEYV IREEWV I✱E WJ U✱N E#2032, 2 0 2 0 J#U0N2 E2 2032,02 0✱
2 0E R✱I CESRSI O
CNS STOEN
C HT N
ECOHL O
NOG LYORGEYV IREEWV I E W 71
9
✱ INTEGRATED ACCESS AND BACKHAUL INTEGRATED ACCESS AND BACKHAUL ✱

References
1. Ericsson Technology Review, 5G New Radio RAN and transport choices that minimize TCO, November 7,
2019, Eriksson, A.C; Forsman, M; Ronkainen, H; Willars, P; Östberg, C, available at: https://www.ericsson.
com/en/reports-and-papers/ericsson-technology-review/articles/5g-nr-ran-and-transport-choices-that- 25 years of experience in this and in system management,
minimize-tco

the authors
area. Throughout his career, as well as heading up the
2. Ericsson Microwave Outlook 2018 report, available at: https://www.ericsson.com/en/reports-and-papers/ he has served in various Algorithm and Simulations
microwave-outlook/reports/2018 leading roles, working on a department at Ericsson
3. Ericsson Mobility Report, Making fixed wireless access a reality, November 2018, available at: wide range of topics. Mobile Platforms/ST-
https://www.ericsson.com/en/mobility-report/articles/fixed-wireless-access Ericsson. Ericsson holds
4. 3GPP TR 36.806, Relay architectures for E-UTRA (LTE-Advanced), available at: a Licentiate Eng. in automatic
https://www.3gpp.org/dynareport/36806.htm control and an M.Sc.
in applied physics and
5. IEEE, Ericsson Research, Integrated Access Backhauled Networks, Teyeb, O; Muhammad, A; Mildh, G;
Henrik Ronkainen electrical engineering from
Dahlman, E; Barac, F; Makki, B, available at: https://arxiv.org/ftp/arxiv/papers/1906/1906.09298.pdf
◆ joined Ericsson in 1989 Linköping University, Sweden.
6. Ericsson Technology Review, 5G evolution: 3GPP releases 16 & 17 overview, March 9, 2020, Peisa, J; to work with software
Persson, P; Parkvall, S; Dahlman, E; Grøvlen, A; Hoymann, C; Gerstenberger, D, available at: https://www. development in telecom
ericsson.com/en/reports-and-papers/ericsson-technology-review/articles/5g-nr-evolution control systems and later
7. 3GPP TS 38.401, NG-RAN; Architecture description, available at: became a software and
https://www.3gpp.org/dynareport/38401.htm system architect for the 2G His current focus is on 5G NR
8. 3GPP TS 38.174, Integrated access and backhaul radio transmission and reception, available at: and 3G RAN systems. With and the strategic evolution of
https://www.3gpp.org/dynareport/38174.htm the introduction of High mobile networks and fixed
Speed Downlink Packet wireless applications. Edstam
9. Ericsson white paper, Advanced antenna systems for 5G networks, von Butovitsch, P; Astely, D; Friberg,
C; Furuskär, A; Göransson, B; Hogan, B; Karlsson, J; Larsson, E, available at: https://www.ericsson.com/en/ Access, he worked as a holds a Ph.D. in physics from
reports-and-papers/white-papers/advanced-antenna-systems-for-5g-networks system architect for 3G and Chalmers University of
4G UE modems. Ronkainen Technology in Gothenburg, Christer Östberg
10. ECC Report 307, Toolbox for the most appropriate synchronisation regulatory framework including
currently serves as a system Sweden. ◆ is an expert in the physical
coexistence of MFCN in 24.25-27.5 GHz in unsynchronised and semi-synchronised mode, March 6, 2020,
designer at Business Area layer of radio access at
available at: https://www.ecodocdb.dk/download/58715ebf-a1e3/ECC%20Report%20307.pdf
Networks, where his work Business Area Networks,
11. 3GPP TS 38.340, NR; Backhaul Adaptation Protocol, available at: focuses on analysis and where he is currently
https://www.3gpp.org/dynareport/38340.htm solutions related to the focusing on analysis and
architecture, deployment solutions related to the
and functionality required by architecture, deployment
The authors would
5G RAN. He holds a B.Sc. in and functionality required by
like to thank
Further reading electrical engineering from 5G RAN. After first joining
Anders Furuskär,
Lund University in Sweden. Ericsson in 1997 to work
❭ Ericsson, Building 5G networks, available at: https://www.ericsson.com/en/5g/5g-networks Jialu Lun,
with algorithm development,
Birgitta Olin,
❭ Ericsson, Microwave backhaul, available at: https://www.ericsson.com/en/networks/trending/hot-topics/
Per Skillermark, Jonas Edstam Anders Ericsson he later became a system
microwave-backhaul ◆ joined Ericsson in 1995 ◆ joined Ericsson in 1999 architect responsible for the
Johan Söder,
❭ Ericsson, Fixed wireless access, available at: https://www.ericsson.com/en/networks/offerings/fixed-wireless- Allan Tart and and currently works with and currently works as a modem parts of 3G and 4G
access Sten Wallin for portfolio management at system designer at Business UE platforms. Östberg holds
❭ Ericsson, 5G access, available at: https://www.ericsson.com/en/networks/offerings/5g their contributions Business Unit Networks. Area Networks. During his an M.Sc. in electrical
to this article. He is also an expert on wire- time at Ericsson, he has engineering from Lund
less backhaul, with more than worked at Ericsson Research University.

72
10 E R I CESRSI O
CNS STOEN
C HT N
ECOHL O
NOG LYORGEYV IREEWV I✱E WJ U✱N E#2032, 2 0 2 0 J#U0N2 E2 2032,02 0✱
2 0E R✱I CESRSI O
CNS STOEN
C HT N
ECOHL O
NOG LYORGEYV IREEWV I E W 73
11
ISSN 0014-0171
284 23-3353 | Uen

© Ericsson AB 2020
Ericsson
SE-164 83 Stockholm, Sweden
Phone: +46 10 719 0000