You are on page 1of 14

CI/CD: ericsson.

com/
5g-core-guide

Continuous software
for continuous change
Your guide to CI/CD evolution for telecom
software in a cloud-native ecosystem

The cloud-native 5G Core network guide series 2.0


2 Ericsson | CI/CD: Continuous software for continuous change

CI/CD today and tomorrow

Continuous integration and continuous deployment (CI/CD)


is the cornerstone of true digital transformation. In this
guide, we’ll share insights into life cycle management (LCM)
automation using CI/CD practices.

Contents Our society is on a journey to digitalization, In our dialogue with service providers,
02 CI/CD today and tomorrow expanding to almost all industries. we have noted different automation
06 Where do we recommend Use cases such as Massive and Critical ambitions and considerations of
service providers start? IoT, public-safety solutions and various how CI/CD will impact their ways
07 CI/CD with Ericsson 5G-enabled services are emerging. This of working. We have found that 5G’s
11 Learnings from global increases the demands placed on your complexity needs to be handled by
CI/CD deployments communication network for features the production system, whether it
13 Summary and higher performance. Perhaps most relies on purpose-built hardware, is
importantly, as digitalization affects executed on off-the-shelf hardware
more and more processes around us, it in virtualized deployment scenarios
requires secure, high-quality connections. or is run in a cloud-native environment.
With this evolution, innovative Likely, all of these will coexist.
communications service providers are CI/CD and automated LCM
asking for speed and agility to launch is particularly expected for the
new features and functionalities in their cloud-native product portfolio, with
networks. A continuous flow of new 5G Core being a prime candidate. But
capabilities and improvements – especially the demand is not limited to 5G products.
with the advent of 5G – is becoming The same requirements exist for
necessary to stay competitive. This is cloud-native implementations
where CI/CD comes in. of the OSS and BSS portfolio.
3 Ericsson | CI/CD: Continuous software for continuous change

The vision for CI/CD in telecom networks Automated handling of updates at more Figure 1: Benefits of CI/CD
The ultimate goal of CI/CD is to allow frequent intervals poses lower risk than in telecom networks
for automated, repeatable and low-risk manual ways of working and large,
updates of all components and layers of infrequent upgrade projects. This allows
the architecture stack. This includes the swift, seamless introduction of the latest Increased efficiency

cloud infrastructure, software applications software release with up-to-date security. • Zero-touch deployment and testing
for the core network and IT domains like Automation in CI/CD pipelines for • Engineering and operations streamline
OSS and BSS, management of network network functions (NFs) is based on ways of working
slices for end-to-end services and the automated flows provided together • Manage new services and increased
traffic more efficiently
exposure of APIs. with the NF and fits into existing, often
To meet service provider requirements, standardized, architectures equally
new releases from suppliers need to be as into operational procedures and Reduced risk
continuously delivered, integrated, verified, service providers’ ways of working.
• Latest software release
validated and deployed. This flow should With all its significant and proven • Remove risks that arise from
be fast, and each step automatically benefits, this automation remains largely manual ways of working
triggered by input from the previous one. independent of the NF’s implementation, • Up-to-date security
• Gradual and frequent updates instead
With updated software made available to either as a VNF or cloud-native application.
of large upgrade projects
service provider repositories, automation To unleash the next level of possible gains,
will pull new and accepted deliveries into the LCM of cloud-native applications must
the next environment, facilitating a flow be treated in the cloud-native way. Service agility
between test and production, or central • Shorter time to market
and national operations. The CI/CD pipeline • Latest capabilities for activation
This will be a journey for the industry Figure 2 illustrates the CI/CD flow from • Ready to address the diversity of
where certain parts still need to be defined a service provider perspective. Suppliers new services
and developed, but one that is necessary to continuously make new software releases
reap the full benefits of 5G and beyond. available, which must then be delivered
Traditionally, CI/CD automation is into the service provider domain. There,
implemented to remove manual operational they are integrated into larger systems,
tasks, achieving efficiency and security configured in line with service provider
benefits as a result. Increased efficiency is needs and then validated against set
achieved when engineering and operations specifications. These activities are
streamline their ways of working and summarized as continuous integration
manage new services and increased traffic. in the lab, test or staging environments.

Figure 2: The CI/CD flow for service providers

Continuous Phased
deployment release to
production site
and slices

Continuous
integration Software testing Validation
and end-to-end testing and monitoring

Customization,
Continuous tuning, integration Service
delivery provider Data
collection

Continuous
Continuous
release
improvement

R&D and
service delivery
Software
supply
Feedback

Continuous
improvement
4 Ericsson | CI/CD: Continuous software for continuous change

Figure 3: Impact on ways of working as more CI/CD automation is added

Use CI/CD and automation acceptance Continuous software delivery to Continuous software delivery
testing in delivery projects customer lab and pre-production to customer production

Phase 1 Phase 2 Phase 3


• Significant transformation of skills • Transformation of skills and ways • Processes for approval and roll-out
and ways of working for vendor of working for service providers in large-scale production network
service delivery • Service provider engineering • Procedures and in-built capabilities
• Mapping of test object lists and creation and operations working as one are designed to handle failure as a
of new cases in test tools • Streamlined processes and trust normal situation
• Coexistence of manual and automatic in automation tools • Defined responsibilities between
acceptance testing • Approval process to mark software service provider, application and
in catalog as ready for production pipeline vendors

New software releases are either rejected in containers, using Kubernetes or going
for further improvement or accepted for to a microservices architecture; it will The three main cloud-native
production at the continuous deployment have a major impact on operations and LCM specific benefits of CI/CD
phase, where they may be upgraded in the LCM on all layers of the architecture in telecom networks are:
service, canary tested and then set for stack. This transformation is not only • speed of change:
normal operation. This process can be about new technology, but is also to – a true continuous delivery model
executed repetitively across sites. a great extent related to new ways of with frequent small updates
A key aspect of CI/CD is applying working and developing required skill – release corrections and new
a DevOps mindset and culture, where sets and processes (see Figure 3). features much more quickly
focus is not only on automation, but We explore the operations aspects in – simple deployment approach
also on feedback loops to identify more detail in our guide “Transforming for pure software assets
any shortcomings early and allow for operations on the way to 5G”, where • openness:
correction. The collection of data from CI/CD is an important part.1 – domain (BSS, OSS, NF) and
the CI/CD flow and the application vendor (inherently multi-vendor)
execution is analyzed through data Moving to a true DevOps mindset independent approach to LCM
analytics feeds into both the software Even though required improvements – separating deployment pipeline
supplier’s and service provider’s to operations are clear for most, many technology from orchestration,
continuous improvement processes. service providers recognize that these including hyperscale cloud
Key steps are executed and tested improvements require a fundamental providers (HCPs)
multiple times in different environments change to how they design and implement – benefit from large CNCF
before deployment in a service provider their business processes. For example, software ecosystems
production network, providing software it means moving beyond today’s ETSI – use modern cloud technologies
quality assurance and a trusted process. MANO-based orchestration and software and HCPs
management which has been mainstream • efficiency:
A phased approach to introduce CI/CD for a decade, to orchestration in concert – independence between
Moving from physical to virtual NFs was with GitOps and Cloud Native Computing functional and infrastructure LCM
a large internal shift but did not have a Foundation (CNCF) tooling and ecosystems. – orchestration and management
major impact on the operational model. CI/CD based on cloud-native LCM is now, to focus on the specifics for
With cloud-native applications on the other step-wise, turning into reality, with some telecom (the NFs, the service)
hand, it is not only about running your code service providers taking bolder steps – replication of configuration
towards its implementation in for troubleshooting purposes
their networks. between environments, including
The time is now for you to further Continuous deployment for cloud-native from the customer to Ericsson
embrace cloud native and start applications can unleash its full potential
adopting new business processes. when implemented using cloud-native
This is because disruptive technology best practices and following the GitOps Cloud-native LCM of applications based
enables cloud native and is maturing, paradigm. Continuous deployment can on microservices, however, cannot be
its values have been proven in IT then draw on the cloud-native LCM implemented as an afterthought on any
and 5G architecture incorporates it. advantages and inherit the benefits of containerized application landscape and
In fact, embracing cloud-native working with version-controlled repositories OSS management stack. This means
architecture and operations is providing a single source of truth, the you must secure applications, and LCM
essential for the secure, efficient possibility to track and manage changes, automation must follow cloud-native best
evolution of your 5G network. and enable stable and reproducible practices with operations adopting suitable
rollbacks to any previous version. ways of working and a DevOps mindset.

1
For more information, please see the Transforming operations on the way to 5G guide.
5 Ericsson | CI/CD: Continuous software for continuous change

Designing 5G applications This means each increment is small and historically bespoke processes for
with automation in mind can easily be verified or rejected, ensuring first-time installation, updates,
CI/CD and LCM for cloud native is a quality and robustness of the main configuration changes or restoration
necessary but insufficient enabler for software branch. are now implemented as a combination
continuous deployment. Prerequisites Following the configuration principle of stopping and instantiation from a
for successful cloud-native LCM include means configuration information is source of truth inside the service provider.
applications being implemented as injected into the runtime environment Rolling updates of the underlying
good tenants in a CNCF ecosystem as variables, or as settings defined in an Kubernetes infrastructure rely on the
(particularly on Kubernetes) and the independent configuration file. The benefit same capability.
adherence to existing guidelines for of separating configuration settings from
cloud-native applications. application logic is that you can apply
This starts with the 12-factor configuration according to the deployment What is GitOps?
application principles on microservice path for lab, pre-production, or multiple, GitOps is a way of working that
level,2 enabling in-service software maybe differing production environments takes best practices for development
upgrade (ISSU) capabilities and evolved in local or HCP clouds. LCM of the and applies them to software
configuration management. It requires configuration (separately from the automation. It uses a declarative
maintenance of a single code base for application) in version-controlled approach that forms the basis
an application to be deployed into many repositories is a prerequisite for of continuous everything.
different environments, as implemented environments and deployments to be
by Ericsson through One-Track for all declaratively described and a GitOps The resulting benefits are:
our applications, and the Application way of working to be implemented. • single source of truth
Development Platform services for code “Disposability” means an application • possibility to track and
which is shared across our portfolio. In the can be started or stopped at a moment’s manage changes
cloud-native paradigm, an application notice. This facilitates fast elastic scaling, • stable and reproducible rollbacks
is composed of microservices with rapid deployment of code or configuration to any previous version
independent life cycles. In CI/CD pipelines, changes, and robustness of production • removing complexity by using
microservices are continuously added deployments. It is necessary for cloud-native best practices
and verified into the main product branch. ISSU and cloud-native LCM, where

2
For further detail, please see 12factor.net.
6 Ericsson | CI/CD: Continuous software for continuous change

Where do we recommend
service providers start?

All basic, repetitive, time-consuming activities that can be


automated must be automated, as they represent the best
potential for savings. We recommend that, as a service
provider, you start with big-ticket items and quick wins.

As efficiency is gained, new automation Figure 4: Examples of automation paths to get started
needs are identified, making the next
steps clear. A typical entry point is
Path 1
the automation of all validation and
verification activities. Additional quick
wins include software download-related Reaching a GitOps state
tasks and selected low-level tasks in
software preparation and deployment. 3 6 9
Functional and software
Examples of possible paths are shown LCM separation
in Figure 4: start with the lab, then move
to production (Path 1); automate tasks
gradually, then take each to production LCM procedure for
(Path 2); or a combination of the two. configuration management

Individual automated steps can 2 5 8


LCM procedure for software
then be orchestrated into larger sections, update and installation
and finally into a CI/CD flow with less
manual intervention.
Service providers’ needs and Automatic acceptance
approaches to CI/CD will vary
depending on their network size 1 4 7
Automated software
and operational and service ambitions. downloads and integrity check
Frontrunners wish to transform their
ways of working and operational
Lab Pre-production Production
procedures to become truly agile in
the marketplace, while continuously
receiving the latest software releases,
even at microservice level. They are Path 2
architecting and implementing CI/CD
based on IT best practices to handle
LCM of products from their suppliers Reaching a GitOps state
in a common way. 7 8 9
Conservative service providers Functional and software
LCM separation
would like CI/CD benefits within their
existing operational procedures and
architecture, for faster product validation LCM procedure for
and quality at reduced cost. They do configuration management
not expect a higher release cadence, 4 5 6
frequent updates or a continuous LCM procedure for software
software flow. update and installation
Whichever road is taken, the
journey to 5G will require a virtualized
network and cloud-native technologies, Automatic acceptance
mandating the selection of applications 1 2 3
designed for automation to ease their Automated software
LCM into a viable CI/CD flow. This downloads and integrity check
signifies a fundamental shift for
service providers. Lab Pre-production Production
7 Ericsson | CI/CD: Continuous software for continuous change

CI/CD with Ericsson

Ericsson’s cloud-native products and solutions are evolving


towards enabling flexible deployment (as opposed to
deployments only based on delivery packaging) and an
“everything-as-code” approach to LCM.

LCM is aligned between our products, Service provider-specific CI/CD All known and unknown products can
as our portfolio implements a common requirements, for example additional test be supported by providing a framework
set of cloud-native principles referring cases and integration of test tools, can be into which product-specific automation
to agnosticity, state-optimized design, accommodated. Our standard delivery is injected. We achieve this by creating
decomposed software, orchestration and deployment flows, as provided by the automation files as part of the design
and automation, and application each product, can be customized to add process and delivering these alongside our
resiliency.3 We have developed a unified functionality or remove individual tasks products, also known as “pipeline-as-code.”
model which includes tools, processes and from the flow. The automation files are then embedded
methods for cloud-native software design: CDD is based on three main principles: in CDD in the target environment. Such
the Application Development Platform. • It is product-agnostic to scale across automation files can also be created offline
This specifies the principles for CI/CD, the entire portfolio. and injected, allowing the opportunity for
provides and maintains a range of generic • It uses selected de facto industry products that have not yet implemented
microservices for consumption in any standards, for example Spinnaker, the automation files, or for third-party
product and delivers tools and methods Ansible and YAML. products, to be orchestrated.
for designing additional, largely • It provides integration points, and For example, in the software upgrade
independent, microservices. builds on tools and products used phase, the LCM procedure for a maiden
With the foundation for cloud-native in the specific target environment. installation or software update is executed,
LCM and CI/CD automation in place, including: pre-update health checks,
you can drive declarative cloud-native This allows the solution to be used and potential traffic evacuation, the software
LCM into your architectures, and evolve configured for ETSI MANO-compliant update and basic post-update health checks.
operational procedures and ways cloud-native network functions (CNFs) The software change for an NF is
of working. and cloud infrastructure, as well as for executed by a software LCM manager;
IT applications that are not managed our product is the Ericsson Orchestrator
Meeting unified and diverse according to ETSI standards. and the Evolved VNF Manager.
service provider expectations
If you subscribe to continuous software
releases of products, it’s natural that you
Figure 5: Simple comparison – traditional vs. future CI/CD
would expect consistent ways of working
in your environment. You will also expect
that available software is automatically
Traditional Future
validated and deployed in your lab and
production networks.
Ericsson CI/CD implementation Modular automation blueprint
based on Continuous Delivery and
Reusable automation assets
Deployment (CDD) automatically updates ”Atomic” automation full stack
(automation jobs)
the service provider lab and production
networks, independent of product domain Reference implementation
or underlying technology. This provides
a uniform way of delivering, preparing, Orchestration in concert with
ETSI MANO management functions
GitOps and CNCF tooling
deploying and validating software.
CDD and the corresponding automation Imperative workflows Declarative system descriptions
workflows are developed, tested
and released alongside the products. CD&D as automation of manual procedures CD&D as an integral functionality
Global usage of CDD in R&D, service
delivery and service providers’ networks Composable pipeline fragments
Long end-to-end use case pipelines
(micro-pipelines)
secures quality for each release.

3
For more information, please see the Cloud-native transformation guide.
8 Ericsson | CI/CD: Continuous software for continuous change

Accelerating LCM Automation is evolving from CI/CD In concrete terms, we clearly separate
LCM is increasing in importance on all being a complementary approach to between LCM of NFs, network-level
levels in service providers’ networks. existing management systems, towards services and systems with service,
LCM of microservices and application being a more integrated CI/CD approach function and resource orchestration
instances in any execution environment using declarative system descriptions as and deployment pipelines for software
does not replace LCM on a logically higher a basis. This allows the movement to a LCM based on evolving cloud-native
NF, network service or system level. In modular automation blueprint with reusable best practices (see Figure 6).
the management, orchestration and automation assets revolving around Within the microservice scope,
assurance domains, focus is shifting from GitOps as a base technology (see Figure 5). focus is on instantiation and upgrade
management of the actual instance in In the automation blueprint, CI/CD of microservices, container images and
an execution environment to securing tools, automation jobs and GitOps are related helm charts including day 0
syntactically and semantically correct used in concert with telecom-focused configuration in any Kubernetes cluster.
representations of the canonical, desired orchestration systems. These systems, LCM is implemented through cloud-native
system state versioned in repositories. based on a network model, determine pipelines, or through better reconciliation
Its task is to determine the desired state the desired system state and store it of a described “wanted state” against
rather than being occupied with the current in a Git repository. The updates in the “running states” in the cluster. The
deployment technology. Git repository then trigger small CI/CD CNCF/IT/HCP ecosystem momentum
Orchestration and assurance on NF, pipeline fragments or GitOps agents, for deploying technology choices is
network service or system level has a which reconcile the desired state with leveraged, and this limits dependency
valuable role to play here. LCM within a target system, such as a Kubernetes deployment automation selections
management and orchestration must, cluster or a CNF. The modular automation in the case of service providers’ own
among other things, ensure it does not blueprint now allows the composition implementations or HCP integration.
break a running service through faulty of various pipeline fragments into larger Pure focus on LCM of microservice
configurations or software updates. pipelines, which are being triggered at and container artifacts facilitates the
Automated measures need to be applied various stages of the automation and same principal approach in all stages
to detect all syntax failures and, as far as orchestration process. of the delivery and deployment into
possible, semantic failures of configuration Our existing CI/CD implementation Kubernetes clusters, even if Ericsson and
files as well as incompatible software approach is evolving to create more service providers take different technology
versions, to reject or trigger corrective transparency and tooling choices for decisions. It enables a common software
actions when possible before applying LCM of the cloud-native software LCM for all domains (OSS, BSS, NFs)
change to a system. The management and artifacts within the CNCF ecosystem, from the implementation perspective.
orchestration suite must take policy and while not losing sight of your needs Running software “gets meaning” and
control responsibility. It can then decide when managing NFs and services is turned into NFs and services by day-1
what is deployed and where to support or their operational automation and day-n configuration, which is adjusted
the required service and populate or needs in lab, pre-production and based on situational awareness, or to
update the runtime repositories. production settings. meet changing needs. It is our view that
network service complexity is solved in
the functional domain, leveraging existing
capabilities of the implementation domain
(cloud native) for execution.

Figure 6: Separation of LCM independence


Network function,

Service orchestration Network service


service scope

Configuration
manager
Function Resource orchestration
Day 1, n
orchestration (Transport network/cloud) Network function
configuration

Configuration
Kubernetes Application
manager
microservice scope

Day 0,
configuration
Container,

Kubernetes Microservice(s)

Docker Container runtime


9 Ericsson | CI/CD: Continuous software for continuous change

Within the NF and services scope, the which hold the representation of which service provider-independent
service provider understands the services software should be deployed and where, procedure, and should be applicable
(business requirements), their intents and and which contain the declared state to cloud-native software from all
has the capability to break these down (day 0 differently handled today from vendors. Reconciliation towards
into the actual resources, applications and day 1, n configuration). the target setting is implemented by
their configurations. The translation of Transparent LCM of the implementation CNCF ecosystem tooling. With tools
network-service intents into configuration on application/microservice level starts like flux evolving, implementing the
of the underlying software can be from these version-controlled declarative desired state in a Kubernetes cluster
supported by service, functional and system descriptions (“as code”). It is is likely to become the common CNCF
resource orchestration as part of the within the cloud-native LCM responsibility way of working.
functional OSS. to resolve the difference between the The CNCF ecosystem is becoming
Our experience has shown a clear declared state stored in the runtime more capable and declarative deployment
separation is required between “upper level” repository and the deployed artifacts pipelines (for example, GitOps) are
functional and service orchestration and running in the Kubernetes clusters. an approach embraced by innovative
“lower level” application and microservice This is an application domain or service providers.
LCM in architectures that implement them
both. Runtime repositories are making a Figure 7: Repositories separating the LCM of functions
responsibility demarcation and decoupling and software artifacts in a Kubernetes cluster
between function and service orchestration
and microservices LCM. Artifact repositories Continuous deployment
are also a demarcation line between of wanted state Phased release
software suppliers and service providers to production site
as shown in Figure 7. and slices
Decoupling between the NF and
microservices scope is implemented Continuous
via runtime repositories (see Figure 7), integration
Software testing Validation
and end-to-end testing and monitoring

Customization, Functional LCM


tuning, integration

Security scanning Declared


deployment
state

Runtime
repositories

Cluster
reconciliation Production
Runtime
Kubernetes cluster
repositories

Artifact Lab/pre-production
repositories cluster reconciliation
Canary testing

Service provider
Continuous Kubernetes cluster
delivery

Data collection
Continuous
release

R&D and
service delivery

Software
supply Continuous
feedback

Continuous
improvement
10 Ericsson | CI/CD: Continuous software for continuous change

From a software-flow perspective, improves organizational performance


artifact repositories are the landing between service provider development, Key aspects of cloud-native
place for artifacts received from vendors. quality assurance and operations. But LCM and CI/CD
Service providers’ internal onboarding instead of maintaining large imperative • Clear separation between the
informs artifact consumers (for example pipelines, these automated flows will LCM of functions and services
orchestration, runtime repository) about be split into micro-pipelines triggered and realization management,
availability of new artifacts and optionally by the service provider or on events the LCM of the underlying
triggers their processing. Integration occurring within delivery, onboarding implementation in terms of
between repositories and control and and cloud-native LCM. applications and microservices.
policy functionality in the management Based on our experience (see Figure 8), • Transparent LCM of
and orchestration domain is needed. we are evolving our CDD toolset to also the implementation of
With artifacts (and potentially adjusted handle these events, and trigger event- and applications/microservices
configurations) being promoted into task-specific micro-pipelines. These can be starts from declarative system
runtime repositories, cloud-native LCM provided by us or bespoke for the service descriptions (“as code”) with the
will reconciliate the cluster and deploy provider. Existing Ericsson automation, for canonical desired system state
the updates. example software onboarding or testing, versioned in repositories.
Traditional automation will still will be able to execute as micro-pipelines • Runtime repositories make a
be needed, and even evolve. The need and as part of a larger imperative flow responsibility demarcation and
for security checks and scans on as needed. decoupling between function
incoming software artifacts, validation and service orchestration and
and verification tasks after cluster microservices LCM.
reconciliation in labs and pre-production • LCM of running microservices
will not disappear or be fully replaced and application instances in any
through canary deployments. Other execution environment does not
customer-specific systems will still replace LCM on a logically higher
need to be triggered. This specific NF, network service or system level.
higher-order internal automation

Figure 8: Ericsson’s CI/CD timeline

Ericsson Customer engagements

Automated Acceptance One-track software introduced CI/CD pipeline industrialized from Architecture for “common CI/CD and LCM
Test introduced in Packet Core development best-practice customer engagement for cloud-native applications on Kubernetes”

RAN CD pilot starts

DevOps introduced GitOps approach defined


for cloud-native LCM

2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022

Agile and one track in RAN RAN: Eight CD – customers 150+ CI/CD
software development including feedback loop engagements worldwide

Software Development System:


CI/CD pipelines on best practice tooling Packet Core CD pilot starts CDD introduced

Ericsson Network Manager CI/CD pipeline RAN CD/CD+ with 20 customers and additionally ~50 customers on a regular
integrated with production systems basis also run pilots in GoToMarket concept using Maintrack software
11 Ericsson | CI/CD: Continuous software for continuous change

Learnings from global


CI/CD deployments

While CI/CD is a new concept for many, we have partners


who are already putting these principles into practice, and
their key learnings can guide your own CI/CD journey.

Evolution of RAN continuous taking pre-commercial deliveries from


deployment (CD) the RAN software provides critical We introduced CD into our network
For RAN, the foundation of the CI/CD feedback about both legacy and new during Q2 2020. It has allowed us
pipeline started in 2012. It focused on functionality and is an integral part of to considerably speed up software
continuous integration and verification, the R&D CI/CD pipeline. validation by trying the functionality
including continuous upgrade capabilities The CD setup serves multiple purposes. in advance, reducing the time
and the automation of test cases in all Service providers benefit from early substantially from commercial
levels of the development flow. In 2015, access to new functionality, giving them release to customer roll-out
T-Mobile US was the first service provider a competitive edge in their markets, compared to earlier releases, even
to run a CD pilot in phase 1 (see Figure 3) and early verification to manage risks while our staff were working remotely
as a proof of concept, taking regular associated with service provider-specific due to COVID-19 lockdowns.
deliveries from the pre-commercial main configurations in small-scale deployments,
track, and carry out release verification in minimizing impact on users. We benefit Elio Iacovacci,
live networks continuously. One year later, from early and continuous feedback on Head of Mobile Access,
CD replaced the earlier release verification how the latest functionality behaves in Telecom Italia
setup, and seven service providers, a live environment at limited scale, a form
including Swisscom, were onboarded for of canary deployment to collect feedback
a long-term partnership with Ericsson, and adjust solutions before releasing to
as described in phase 3 (Figure 3), a global audience. The CI/CD pipeline CD is key for us to win the network
continuously deploying bi-weekly, has also proven to be an enabler for performance benchmarks, and has
pre-commercial software into their more frequent commercial releases. given us the confidence to deploy
commercial networks. This replaced the In order to scale the impact of the service RAN software nationwide in just two
previous release verification setup, which provider benefits described above, the days. The total effort associated with
largely corresponds to the continuous intention is to increase the number of network upgrades is similar to before,
integration in Figure 2, thereby collapsing RAN CD customers over the coming years. but is now spent earlier and is more
this component to manual or automated This will also pave the way for service evenly spread in time, enabling more
software distribution. As of 2022, Ericsson provider operational transformations efficient planning and the possibility
RAN collaborates closely on CD with and automated pipelines required for to continuously allocate a small
22 service providers. materializing the full potential of the team instead of organizing
Through continuous automatic data cloud RAN deployments of the future. major, infrequent projects.
collection from the CD zones, the feedback The high number of customers
loop is closed, enabling insights and onboard with RAN CD will make it easier Mark Düsener,
resolution of issues from the software main for you to introduce cloud RAN where Head of Mobile Network,
track deployed in commercial networks. new opportunities with CI/CD concepts Swisscom
Close collaboration with CD customers described in this paper will be available.
12 Ericsson | CI/CD: Continuous software for continuous change

Core networks Many service providers we work with


Core networks, and the cloud-native have, like us, identified that they need The 5G Core comes on top
5G Core in particular, are also domains new ways of working and need to change of containers, which is a new
that we collaborated with service providers the ways they work with vendors to realize technology that is enabling new
on early to explore the CI/CD concept. a continuous delivery model. ways of delivery. It’s about
Building on Packet Core one-track A lack of trust in software continuous implementation and
software development and CD pilots in updates/upgrades in the past led to continuous delivery, and both
2015, the journey has continued with a long processes, with many interactions Vodafone and Ericsson needed
common CI/CD framework addressing proving hard to automate and efficiency to learn how to adapt to a new
the main automation challenges for difficult to achieve. While the modelling environment and a new delivery
ETSI MANO-compliant CNFs, as well as of existing processes into pipelines method. We selected one of our data
for IT applications that are not managed seemed a natural first step and helps centers and, together with Ericsson,
according to ETSI standards. to reduce impact on organizations, we implemented the 5G Core directly
Learnings from service provider it can prevent operational gains. into the live system, which is a
engagements globally show that Some service providers keep the brand-new approach. This required
automation developed and delivered CI/CD automation processes and tools a change of mindset in terms of
together with a product allows easy for cloud-native and legacy software how we bring this new system
adoption. CI/CD pipelines for the separate. This allows them to build up to life, but also how we bring
preparation, deployment and validation new teams and ways of working, fully the new service to life in a live
of new software versions is best done by leveraging new technologies without network environment.
the vendor that designed the software. impacting legacy staff and processes.
Service providers are looking to The teams managing the new technologies Guido Weissbrich,
orchestrate these larger automation flows selectively pick best practices from Chief Network Officer,
in a pipeline structure representing their established ways of working. Using Vodafone Germany 4
ways of working and organizational setup. an architecture with micro-pipelines
This is the most pragmatic approach, facilitates the selection and reuse of
allowing service providers to focus their existing best practice. It is important to
development on what is specific to their protect current investments and assure By taking a greenfield approach to
environment. Going forward, we observe quality for critical infrastructure when deploying its cloud-native dual-mode
a demand for even smaller task-specific moving to cloud native. Software vendors 5G Core, Telstra had more freedom
“micro-pipelines” to give service providers work in a similar way, building new to try out new ideas with respect
even more flexibility. It facilitates DevOps teams to manage cloud-native to things like cloud-native toolsets,
integration with service provider-specific products, while leaving the current CI/CD pipelines, test automation
CI/CD technology selection and cloud structure in place for legacy. and continuous deployment
platforms (HCPs) that require delegated The experiences of the first service activities. With cloud native,
and clearly separated responsibilities providers to transition from integrated Telstra now has the ability to
between orchestration, products and solution level automation towards LCM deploy software to deliver new
platform vendors. on individual products and microservices features in a very intuitive manner.
In current customer engagements show that CI/CD is best solved as a
and dialogs, we see a split between combination of orchestration and David Aders,
service providers continuing an ETSI NFV best practice LCM procedures for Group Owner for Mobile
approach and others driving declarative software artifacts. Development & Product
cloud-native LCM into their architectures, The design of micro-pipelines should Engineering,
sometimes competing within a service be based on open source; however, Telstra5
provider. Some are looking at a 2023 security, integration and usability must
implementation horizon. In the latter be added on top to allow integration to a
case, service provider RFQ requirements, service provider production environment.
open-source activities and internal proof This is an R&D activity requiring
of concepts demonstrate that today’s understanding of critical infrastructure
LCM of Ericsson 5G Core on CCD in the service provider domain and the
Kubernetes clusters are in line with micro-pipelines should be validated
the architecture described above. together with the software.

4
For the full Vodafone Germany case study, please see: Shaping Germany’s 5G future.
5
For the full Telstra case study, please see: Delivering 5G in Australia.
13 Ericsson | CI/CD: Continuous software for continuous change

Summary

Embracing automated software pipelines is not optional – it’s a must.


The more you follow cloud-native best practices and standardized,
common ways of working, the closer you’ll move to the end goal for
all service providers – a zero-touch, automated pipeline.

The software and IT industry continues We’ve identified five key actions service The implementation of CI/CD is a journey,
to evolve at an ever-increasing pace. The providers must take as their next steps: which has to be addressed in a stepwise
telecom industry has been leveraging this 1. Separate the functional management approach. It is essential that each service
momentum and scale by continuously from the software LCM, enabling a provider takes a holistic approach to their
adopting IT approaches. We understand common approach for software LCM specific journey, based on their strategic
that, as a service provider, you have for all domains priorities. It’s clear that the underlying
a persistent need to reduce costs and 2. Evolve the software LCM towards technology plays an important role and is
get new updates, functionality and a declarative approach, that can be evolving as IT technology is being applied
products to market quicker. This drives consistent for different infrastructures in telecoms. However, the transformation
the need for an evolution of the approach including HCPs requires much more. It needs development
for CI/CD capabilities, both in terms 3. Evolve orchestration capabilities to of new processes, ways of working and
of technology but more importantly in leverage the declarative software LCM competencies, to name a few.
terms of your processes, ways of working approach and focus it on functional LCM It is our strong recommendation that,
and the development of new skill sets 4. Evolve the delivery pipeline to beyond whatever stage you are at on your journey,
and competences. monolithic deliveries to smaller artifacts you start planning your transformation
This process must be started now. such as microservices project today. We are happy to share our
With the implementation of cloud native, 5. Engage in an automatic feedback experience and learnings from being early
our industry will, to a much larger extent, loop with data from the service to market with leading service providers
compete for competence with other tech provider to Ericsson globally, and look forward to helping you
industries, where programs for upskilling shape your transformation journey and
the existing workforce will play a crucial Although IT technology has matured develop impactful strategies.
role to ensure competence needs are met. to a level where it is applied in telecoms,
We’ve learned this from our experience we still see the need for further alignment
collaborating with service providers, for in the industry. To facilitate these changes, This guide is part of our cloud-native
example our work with Singtel on one of we are actively participating in different 5G Core network guide series 2.0,
the world’s first commercial 5G standalone industry forums to improve areas such as: which explores the topics that should
deployments, where their key learnings and • creating a consistent approach for how be considered when deploying and
insights were characterized as the three to structure the repositories, reducing evolving your 5G Core network.
Ps: people, processes and platform.6 integration costs To discover the full series, please visit
Service providers expect CI/CD to • forming a technology-agnostic approach ericsson.com/5g-core-guide.
transform its current manual processes, to describing the wanted state in the
from a process that takes 90 days to declarative repositories (today it is
introduce software upgrades once, twice, specific to each technology such as flux)
or at most 4 times a year (per NF), into
a process that takes 3 days to get an
upgrade into production. This will enable
the continuous deployment of software
across networks. The resulting agility
and efficiency benefits will translate
into improved customer experience
and value creation.

6
For the full Singtel case study, please see: Transforming Singapore’s future with 5G.
About Ericsson Ericsson enables communications service providers
and enterprises to capture the full value of connectivity.
The company’s portfolio spans the following business
areas: Networks, Cloud Software and Services, Enterprise
Wireless Solutions, Global Communications Platform,
and Technologies and New Businesses. It is designed to
help our customers go digital, increase efficiency and find
new revenue streams. Ericsson’s innovation investments
have delivered the benefits of mobility and mobile
broadband to billions of people globally. Ericsson stock
is listed on Nasdaq Stockholm and on Nasdaq New York.
www.ericsson.com

Ericsson The content of this document is subject 25/287 01- FGB 101 0909
SE-164 80 Stockholm, Sweden to revision without notice due to © Ericsson 2022
Telephone +46 10 719 0000 continued progress in methodology,
www.ericsson.com design and manufacturing. Ericsson
shall have no liability for any error or
damage of any kind resulting from the
use of this document

You might also like