1 of 152

Nephio R4 Developer Summit - 2024 (Day 1)

2 of 152

LFN Welcome!

Arpit Joshipura

GM/SVP, Networking + Edge/IoT, The Linux Foundation

Jenn Bonner��Sr. Program Manager, The Linux Foundation

3 of 152

Logistics items

  • No photography outside of the room (strictly prohibited).
  • Do not leave the room without a Google representative accompanying you.
  • The area outside the conference room is a Googler workspace. Please maintain silence.
  • Restrooms are located at [location].
  • Food will be served in the room.
  • Happy hour will begin right after the summit (around 5:00 PM).
  • Please drink responsibly.
  • Google and LF are not responsible for any liabilities.
  • Please adhere to LF policies.

4 of 152

Nephio Momentum

https://docs.google.com/document/d/1yGCKM1wvGkGLAkIb8ZcjC1ND8OmtYSWXMHM1obj_DMc/edit

5 of 152

Nephio & OSS Networking: Where We Are

! 3nncCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCC Q111111111111111111`````````````````````````````

SMB/Store

Residential

Mobile

Service Provider Edge

User Edge

Carrier

Access

Enterprise & IIOT

Carrier Core

Apps/Internet /

Web

Public Cloud

Enterprise

Core & Cloud

Private Cloud

eg Google, Microsoft, AWS, IBM, Huawei, Alibaba, Baidu, Tencent..

NW API, Functions, Workloads, Life Cycle Mgmt

MANAGEMENT

ORCHESTRATION & ANALYTICS

NETWORK �CONTROL

INFRASTRUCTURE

xNFs

* Sample projects only

Network OS &

Data Plane

Last Mile Enhanced Access

6 of 152

Nephio TSC Welcome!

Sana Tariq

Principal Architect,

Telus

TSC Vice-Chair, Nephio

Kandan Kathirvel

Group Manager, Google Cloud

TSC Chair, Nephio

7 of 152

Goals of this Development summit - R4

Discover Nephio

  • Overview of Nephio, concepts and objectives for newcomers

Plan for R4

  • Focused community discussion to make progress towards R4

Build a Community

  • TSC & SIG get together: Plan for R4 and beyond

8 of 152

We started with this Mission!

Nephio’s goal is to deliver carrier-grade, simple, open, Kubernetes-based cloud-native intent automation and common automation templates that materially simplify the deployment and management of multi-vendor cloud infrastructure and network functions across large scale edge deployments. Nephio enables faster onboarding of network functions to production including provisioning of underlying cloud infrastructure with a true cloud native approach, and reduces costs of adoption of cloud and network infrastructure.

9 of 152

Nephio fundamentals

Intent object describes the desired state

APIs & Data Models

AI/ML Analytics

Network Orchestration

Intent

Intent objects validated and compiled

Network state reconciled with intent

Continuous visibility of network state

Network is constantly self-optimized

10 of 152

Solves the real problem!

Efficient Multi-cloud Infrastructure and Multi-vendor Network Function Configuration Management

Telco Network Automation

Cloud Infra Resources

Automation (Public & Private)

End-to-End Service Orchestration + End-to-End Slice Management

Workload Resource

Automation (NFs)

Workload �Configuration (NFs)

K8s Operator

K8s Operator

K8s Operator

CRD

CRD

CRD

CNF

CNF

Kubernetes Engine

Cloud Infrastructure

Open APIs

K8S Custom Resource Definitions (CRD) driven API

Configuration blueprints for network functions (multiple vendors/multiple domains) and corresponding infrastructure

Operator implementation translating intent to actual configuration, continuous observation and reconciliation on deviations

Applying Kubernetes Resource Model principles

This unified approach allows for the use of CI/CD principles and common tooling across all layers.

11 of 152

Nephio community trajectory

12 of 152

A 2+ Years with Nephio, market feedback and new trends

Use cases &

Verticals

  • Infra, NF, configurations
  • RAN + Core + Transport
  • + Enterprise use cases

Project

Synergies

  • O-RAN, ETSI, TMForum
  • Project Sylva & CAmara
  • LF AI & Data

Evolving

Technologies

  • Kubernetes KRM
  • SDK extensibility
  • + Data & AI transformation

Nephio

Adoption

  • Great market feedback
  • Building good momentum
  • + Lighthouse deployment(s)

13 of 152

Nephio: Expanding Beyond Telecom

The Automation Challenge: Large-scale edge cloud deployments, whether connected or disconnected, demand robust automation across all sectors.

Nephio's Advantage: Provides a vendor-neutral, open-source platform for managing complex cloud infrastructure.

Cross-Industry Applicability:

  • Retail: Streamline operations, enhance experiences.
  • Enterprises: Modernize IT, boost efficiency.
  • FSI: Secure finance, automate processes.
  • Government: Optimize services, improve efficiency.
  • Other industry

Community-Driven Growth: Collaborate to extend Nephio's capabilities and make it the go-to solution for cloud-native automation across all industries.

14 of 152

Community: Things to consider

  1. Tooling Completion: Prioritize finishing core tooling to provide a solid foundation for users.
  2. Quality over Quantity: Shift emphasis from rapid feature additions to ensuring existing features are robust and reliable.
  3. Multi-Cloud & workload automation: Accelerate efforts to deliver open CRDs and configuration capabilities for true multi-cloud orchestration.
  4. Community Growth: Actively engage developers, students, and freelancers. Recognize and reward significant individual contributions (10K to 5K community awards).
  5. AI-First Approach: Expand Nephio's capabilities to optimize and simplify AI workload deployment, addressing a crucial market need.

15 of 152

Network Automation Challenges for Telco

1

2

Heterogeneity and complexity of Network Functions information exchange models

Heterogeneity and complexity of Telco Clouds in Infrastructure Configuration and Automation

3

Heterogeneity and complexity of NF configuration

Nephio attempts to solve these by adoption of ‘common’ Kubernetes Models

16 of 152

How Nephio Solves These Challenges

Nephio R1

  • Kubernetes-based Automation
  • Intent-Driven Management
  • Package Specialization
  • Network Function Automation (5G Core)
  • UI Improvements

Nephio R2

  • Multi Vendor Support
  • Multi Cloud Support
  • Network Function Automation (O-RAN)
  • GitOps support (Porch)
  • Improved UI, Website and Documentation

Nephio Experimental

  • Nephio SDK
  • Helm to operator convertors
  • Flux based Helm deployments
  • Topology Controller

Nephio R4 and Beyond

Readiness for Production

Multi-cloud

Security

Network Functions and Use-Cases

Support for NB APIs (SO, O-RAN SMO …)

Support for NF and NS lifecycle

Synergy with OS and Standards (Sylva, ORAN …)

Nephio User Experience

Enhanced UI, Design Studio

GenAI for automation simplification and LCM Operators SDKs (Helm, Yang/TOSCA )

Nephio R3

  • Functional robustness and Core evolution
  • Network Function Automation (O-RAN)
  • GitOps support (Porch)
  • Improved UI, Website and Documentation
  • OpenSSF score improved to 7.6

17 of 152

Looking Ahead: Nephio R4+ O-RAN Progress

SMO and RAN Domain Management

  • Deploy Near-RT RIC without SMO
  • Deploy xApp without SMO

O-RAN NF

  • NF Orchestration Using SMO
  • Reconfigure O-RAN NF

O-Cloud

  • Create O-Cloud K8s Cluster
  • Delete O-Cloud K8s Cluster
  • O-Cloud Registration

In Progress

Ready for development

18 of 152

Looking Ahead: Nephio R4+ Platform Evolution

Platform and Infrastructure

  • Make Nephio work on bare metal clusters
  • Support for multiple Git and GitOps
  • Core Technology Evolution

Security

  • Increase OpenSSF score to 8 (GitHub Issue)
  • Identity and access management
  • Nephio IaC scanning
  • Workload Identity

Service Assurance and Observability

  • Observability for workloads and platform
  • Closed loop service assurance

User Experience, Modeling, and GenAI

  • Nephio SDK
  • Information and Data models for Nephio
  • RAG Pipeline for GenAI based templates creation

19 of 152

Nephio SIG Structure and Roles

SIG1 Architecture

Workgroups:

  • WG1: Architecture
  • WG2: O-RAN
  • WG3: Rx Scope
  • WG4: Modeling and APIs
  • WG5: Service Assurance
  • WG6: GenAI
  • WG7: Transport

SIG3 Release

Work areas:

  • Improve release machinery
  • CI/CD testing
  • GitHub integration
  • Git branching and tagging
  • User documentation
  • Compile binaries (Go, Makefiles)
  • Build container images (Docker Hub)

SIG4 Security

Work areas:

  • Workload/service identity
  • Infrastructure security
  • Apply OSSF best practices
  • Secrets management
  • Security response team

SIG2 Automation

Team project boards:

20 of 152

Realizing Nephio Business Value for Telcos

Panel Discussion

21 of 152

Realizing Nephio Business Value for Telcos

Daniel Freiji

Director Telco Cloud

Telus Canada

Bernard Tsai

Lead Architect

Deutsche Telekom

Germany

David Kinsey

Lead Principal Architect, ATT, USA

Moderator

Sana Tariq - Telus

Eric Debau

Head of R&D

Orange, France

22 of 152

Cloud Giants Weigh In: Perspectives on Nephio's Network Automation

Panel Discussion

23 of 152

Cloud Giants Weigh In: Perspectives on Nephio's Network Automation

Tomas Fredberg

Senior Expert in IT & Telco Data Center Architectures, Ericsson

Moderator

Jenn Bonner - LFN

Balaji Balasubramanian

Engineering Lead, Google

Bill Wright

Edge and AI specialist, Red Hat

Tom Nadeau

Chief Architect of Solutions & Innovation, WindRiver

Akshatha Sathyanarayan

Director, R&D Management, VMware by Broadcom

24 of 152

Nephio Adoption: Network Function Vendors Navigate the Cloud Native

Panel Discussion

25 of 152

Nephio Adoption: Network Function Vendors Navigate the Cloud Native

Ciaran Johnston

Senior Expert in OSS & Programmable Network Architectures, Ericsson

Moderator

Kandan Kathirvel - Google

Jane Shen

VP Technology Strategy, Mavenir

Wim Henderickx

Head of architecture & technology IP, Nokia

Sundar Nadathur

Principal Engineer, Intel

Avinash Bhat

Director, Samsung Electronics

26 of 152

Open Source Synergy: Nephio's Collaborative Ecosystem with CAMARA, O-RAN, Sylva, OpenAirInterface, and Kubenet

Panel Discussion

27 of 152

Open Source Synergy: Nephio's Collaborative Ecosystem with CAMARA, O-RAN, Sylva, OAI, and Kubenet

Moderator

Timo Perälä Nokia

Eric Debau

Head of R&D

Orange

Sylva

Bill Wright�Edge and AI specialist�Red Hat�CAMARA

Sagar Arora

Solutions Architect

OAI Software Alliance

OAI

Wim Henderickx

Head of Technology and Architecture

Nokia

Kubenet

Seshu Kumar Mudiganti

Principal Technologist

WindRiver

O-RAN SC

Moderator

Sana Tariq - Telus

28 of 152

Welcome to the Nephio Project

Nephio Architecture, Principles, Technology, and Evolution

Nephio SIG Leads

29 of 152

Nephio’s Evolving Architecture

NFVO/Core Orchestration

ORAN Orchestration (SMO)

CRDs

GitOps/CI/CD

Service Orchestration

Infrastructure

Network Fabric

Cloud

Nephio Core APIs

Resources

Inventory

Infrastructure

Config

Workload

Nephio Core Controllers

Token

Bootstrap pkg

K8 Cluster

Repository

Network

Templates

SDKs

30 of 152

Nephio R4 Architecture

31 of 152

Development Process

  • Release every 6 months
  • Mix of waterfall + agile development approaches
  • SIG NetArch and SIG Security develop user stories
  • Hand-off to SIG Automation and SIG Release
  • Breakdown into development tasks (on GitHub)
    • Coding
    • Scripting (builds, tests)
    • And a lot of testing!
  • Feature freeze
  • Code freeze

32 of 152

Nephio Release Process

33 of 152

O-RAN Automation Nephio R4

Joey, Vish, Bala, Stefan

34 of 152

O-RAN Working Group Objectives

Specification of use cases and APIs

Implementation of use cases and APIs

Working group aims to accelerate standards and make Nephio the de facto reference implementation of O-RAN specifications

35 of 152

O-RAN WG2 R4 User Stories

Ongoing:

  • Create O-Cloud Node Cluster
  • NF Orchestration Using SMO
    • Deployment
    • Termination
  • O-Cloud Registration

Backlog:

  • Cluster Registration
  • Delete O-Cloud K8s Cluster
  • SMO Configures O-RAN NF
  • FOCOM Inventory Subscription

36 of 152

O-Cloud Registration

  • In early stages of story grooming and scope discussions.
    • Scoping and implementation to span multiple releases
  • Reference ->

https://docs.google.com/document/d/1g7Z9XIGnBAXnwvISgS69maMXzifsxpJsKJu28lC_PyU/edit#heading=h.mx9z3lrinmyt

37 of 152

O-Cloud Registration

38 of 152

O-Cloud Registration

39 of 152

O-RAN Integration Architecture for Nephio

The Nephio O-RAN WG has agreed upon an Nephio based O-RAN integration architecture

One key aspect of this architecture is that the SMO FOCOM/NFO and the O-Cloud IMS will be implemented with separate Nephio instances / management clusters

The integration between the two Nephio instances will be done only through the O2ims interface

This will provide an SMO/O-Cloud vendor decoupling and also enable non-Nephio based implementations in either the SMO or the O-Cloud

40 of 152

Federated O-Cloud Orchestration

The role for the SMO FOCOM service is to provide federated infrastructure orchestration across multiple O-Clouds

It will integrate towards multiple O-Cloud IMS systems over their O2ims interfaces

Each O-Cloud IMS will expose its site/HW infrastructure through the O2ims Inventory API and support K8s cluster LCM through the O2ims Provisioning API

41 of 152

O-RAN O-Cloud Template

In the O-RAN WG6 O2 Interface General Aspects and Principles specification the concept of an “O-Cloud Template” has been introduced

  • Abstracts the O-Cloud implementation specific artefacts and HW configuration from the SMO�
  • Exposes only high-level characteristics and capacity required for the SMO Service Orchestrator to select a suitable cluster template based on the Cloudified NF/NF Deployment requirements

The template characteristics will also be used to find O-Cloud Sites with matching characteristics and capacity as part of Service Orchestration homing decisions

42 of 152

O-Cloud Templates in O-Cloud IMS and SMO FOCOM

Both a Nephio based FOCOM and a Nephio based IMS will use the KPT/Porch based cluster package management solution implemented in Nephio R1/R2/R3

  • The “Cluster template list” on the SMO side in the O-RAN specification will be realized by a Git based template blueprint repo where the KPT packages for the O-Cloud Templates are onboarded�
  • The SMO-side O-Cloud Template blueprint packages will refer one-to-one to a corresponding IMS-side O-Cloud Template in a specific O-Cloud�
  • If the O-Cloud IMS also is Nephio based, then it will have a Git template blueprint repo as well where the O-Cloud Templates are stored as KPT packages, while for non-Nephio based IMS the O-Cloud Templates may be stored in any persistent database

The IMS-side O-Cloud Template blueprint package is expected to consist of a common part with O-RAN standardized information that will be shared between the O-Cloud IMS and SMO FOCOM, and an implementation specific part with O-Cloud specific artifacts

  • The SMO must be able to query for the common section of the template in order to automatically create the SMO-side O-Cloud Template blueprint package

The SMO-side O-Cloud Template blueprint package will contain the same shared O-RAN standardized properties plus its own FOCOM specific artefacts for southbound integration over O2ims

43 of 152

Nephio FOCOM and IMS integration

The O2ims provisioning interface exposed by the IMS Nephio implementation will provide a declarative NBI towards FOCOM, supporting a “ProvisioningRequest” CR that is aligned with the O-RAN WG6 O2ims provisioning standardization discussions

The FOCOM Nephio implementation must provide a corresponding declarative NBI towards its clients

  • The proposal is to support a “FocomProvisioningRequest CR with partly the same properties as the O2ims ProvisioningRequest CR�
  • The FOCOM NBI will be Git/Porch based and support GitOps based deployment and version handling of the FocomProvisioningRequest CR

44 of 152

Detailed Architecture with Nephio Components: NF Deployment LCM

45 of 152

Detailed Architecture with Nephio Components: Cluster LCM

46 of 152

Cluster Creation User Story implementation scope for R4

  • Umbrella Nephio issue: User Story: Create O-Cloud Node Cluster · Issue #760 · nephio-project/nephio · GitHub
  • Proposed IMS scope for R4 are tasks 1 and 4
  • Proposed FOCOM scope for R4 are tasks 1, 3 and 4
  • Latest O-RAN O2ims provisioning interface modelling proposals shall be considered
    • Define an O2ims “ProvisioningRequest” CRD instead of a “ClusterRequest” CRD
  • Excluded tasks in R4:
    • CRD status monitoring (IMS and FOCOM task 2)
    • Workload cluster credentials handling and exposure to FOCOM (IMS taks 3)
  • Other aspects simplified in R4
    • No automated workload cluster registration in NFO (connected to that credentials are not exposed)
    • No new “ClusterClaim” CRD introduced on the IMS side
    • Use the existing Nephio workload cluster blueprint as the O-Cloud Template on the IMS side

47 of 152

Nephio R4 IMS scope (1/2)

SMO/FOCOM shall only be exposed to the O2ims standardized ProvisioningRequest CR

The Nephio IMS implementation shall show how the information in the O-RAN O2ims CR is mapped into the Nephio internal workload cluster package deployment

For R4 it is proposed that the existing “nephio-workload-cluster” blueprint is used as the O-Cloud Template

The O2ims ProvisioningRequest CR will contain the following input parameters:

  • metadata.name: In O-RAN this is now defined as the SMO/FOCOM-assigned “ProvisioningRequestID”
  • name: Human readable name of the Provisioning Request
  • description: Description of the Provisioning Request
  • templateName: Name of O-Cloud Template
  • templateVersion: Version of the O-Cloud Template��Proposal for Nephio R4: The O2ims operator constructs the Nephio blueprint name as <templateName>_<templateVersion>. The “templateVersion” part shall not be the blueprint package revision, it shall be part of the blueprint package name in Nephio

  • templateParameters: Instance specific input data��Proposal for Nephio R4 for instance specific input:�Include one “nodeClusterName” parameter that the O2ims operator will use as the name for the cluster package instance in the Cluster mgmt repo. This will allow the SMO to use a separate unique identifier for the ProvisioningRequestId in the metadata.name parameter.�� Include one “labels” parameter that the O2ims operator will use to assign labels to the KPT package instance manifest files.� E.g.: “nephio.org/site-type: edge”

The “Nephio-workload-cluster” contains manifest files for the Nephio WorkloadCluster and Porch Package Variants

  • The PV for the CAPI Kind cluster creation can be directly reused
  • All the PVs for creating the mgmt-staging resources (e.g. ConfigSync) can be reused, it will be the responsibility of the IMS to deploy these into the workload cluster
  • The PV for Git and Porch repo creation can potentially be removed, these repos shall instead be created in the SMO FOCOM/NFO Nephio instance as part of Cluster registration

48 of 152

Nephio R4 IMS scope (2/2)

The Cluster1 package deployment will trigger the normal cluster creation workflow in the IMS Nephio management cluster

ConfigSync will deploy the applicable Package Variant CRs in the IMS management cluster that in turn will clone the relevant example packages into the mgmt and mgmt-staging repo

The IMS Nephio management cluster will use internal IMS specific logic to do the actual cluster deployment. This is illustrated here by the “Cluster1 IMS internal Cluster PV” that will clone the package containing the Cluster1 IMS specific CR manifest files

  • In Nephio R1 and R2 the cluster-capi-kind example package has been used to demonstrate this internal logic using CAPI
  • For the Nephio R4 implementation the proposal is to use the same cluster-capi-kind example package

The packages deployed in the mgmt-staging repo will trigger the Cluster Bootstrap Controller to deploy the included manifests into the new Cluster1 workload cluster

  • In O-RAN the installation of any additional cluster-wide software components should be done by the IMS and not the FOCOM/NFO
  • The required software components (ConfigSync CRD/operator, CNI drivers etc.) shall be defined in the O-Cloud cluster template
  • Thus the IMS Nephio management cluster shall realize the Cluster Bootstrap Controller
  • The Cluster Bootstrap Controller will authenticate towards the workload cluster with admin account credentials, fetched from a secret stored in the IMS Nephio management cluster. These credentials shall not be exposed to SMO FOCOM/NFO.
  • The creation of the credentials required for the NFO O2dms authentication is not in scope of the R4 implementation

49 of 152

Nephio R4 FOCOM scope (1/2)

Each O-Cloud shall first be registered into FOCOM through an “OcloudRegistration” CR

  • This CR specifies the IMS endpoint URI and a reference to a secret stored in the FOCOM management cluster with the IMS authentication credentials
  • In Nephio R4 this will be a Nephio proprietary method for O-Cloud registration

The SMO-level O-Cloud cluster template KPT package is proposed to contain two manifest files with Nephio-defined CRDs

A “TemplateInfo” CRD with the following parameters:

  • “oCloudRef” referring to the name of the OcloudRegistration CR
  • “name” and “version” of the O-Cloud Template
  • “templateParameterSchema” that is a JSON schema to be used for machine validation of the clusterTemplateInput
  • “characteristics” and “metadata” as defined by the O-RAN standard

A “FocomProvisioningRequest” CRD with the following parameters:

  • “oCloudRef” referring to the O-Cloud that owns the referred IMS-level O-Cloud cluster template, i.e. the name of the OcloudRegistration CR
  • “name”, human readable name of the Provisioning Request
  • “description”, description of the Provisioning Request
  • “templateName” and “templateVersion” referring to the O-Cloud Template
  • “templateParameters” carrying cluster instance specific input data

Creation of a new cluster is triggered by cloning an O-Cloud Template package blueprint into a Cluster package deployment and updating it with cluster instance specific input

50 of 152

Nephio R4 FOCOM scope (2/2)

As part of the cluster instance package clone/update/approval process there could be KPT specializer functions that are invoked in order to validate the deployment request

  • This can be introduced as a long-term feature but should be excluded from the Nephio R4 implementation
  • Validation of user input can be done by using the templateParameterSchema
  • Validation of the target site can be done by doing a lookup of the O-Cloud site and underlying HW resources through the O2ims inventory API and compare the HW type capabilities with the template characteristics requirements

When the new cluster instance package has been approved, the ConfigSync operator will reconcile the FocomProvisioningRequest CR into the etcd database

The FocomProvisioningRequest CR is managed by a FOCOM O2ims Operator controller logic

  • The referred O-Cloud CR is fetched from the OcloudRegistration CR to get the O2ims endpoint and credentials
  • The FocomProvisioningRequest CR is transformed to the O2ims ProvisioningRequest CR
  • FOCOM submits the O2ims ProvisioningRequest CR to the O-Cloud1 IMS

Outside the scope of Nephio R4:

  • After creating the ProvisioningRequest CR, FOCOM will start monitoring the status of this CR instance and update the FocomProvisioningRequest CR with the reported information
  • Once the ProvisioningRequest CR is reporting status “Ready”, FOCOM will proceed with the registration of the new cluster in the NFO

51 of 152

Kubenet

52 of 152

Kubenet

Initiative focussed to help network engineers understand the potential of Kubernetes principles for network automation/orchestration

Open source projects supporting the initiative

53 of 152

Kubenet: Use cases

- Data Center networking

- WAN networking

- Peering

- Access and campus networking

- Core networking

- Backhaul/Fronthaul

- Cloud Networking

54 of 152

Kubenet: High Level perspective

Resources

Network�Design

Network

Business�Logic

Abstract

DeviceConfig

  • Node
  • Link
  • IP
  • VLAN
  • AS
  • Addressing
  • Protocols
  • Encapsulation

Abstract

DeviceConfig

Vendor

DeviceConfig

Vendor

DeviceConfig

55 of 152

Kubenet: Resources

Node

Link

Port

Endpoint

Adaptor

Module

ModuleBay

IPIndex

IPClaim

ASIndex

ASClaim

VLANIndex

VLANClaim

Network

Topology

NetworkDesign

NodeTemplate

Interface

SubInterface

NetworkInstance

BGP

BGPNeighbor

BGPDynNeighbor

RoutingPolicy

56 of 152

Kubenet: Supporting projects

Cloud native YANG

KRM based inventory and identity management (IPAM, VLAN, AS, etc)

KRM orchestration: Kform/Choreo

KRM Package Manager

57 of 152

Want to learn more, join us

58 of 152

Nephio Core Evolution

59 of 152

Nephio Core Evolution: The Path Forward

Problem Statement

Nephio’s Core technology relies on Porch for package management, manipulation, and lifecycle operations, i.e. package orchestration. This core technology is important for Nephio’s narrative supporting declarative models, specialization and distributed actuation. Nephio community observed challenges with Porch maturity, scalability, and stability, hence there is a need to explore and evaluate or just enhance existing core technologies that can realize Nephio requirements and vision and offer a stable development opportunity to focus on Nephio’s mission to mature cloud native Kubernetes based NF orchestration.

We have looked into 4 technologies: Ansible, TKO, Kubenet, Porch and possibly modified Porch. We also intend to look into Argo CD and Flux CD in the near future.

60 of 152

Criteria for Comparison

Package Management Capabilities

Ability to define, version, and manage KRM-based application packages

Integrations, Developers and Adoption

Ability to work with Kubernetes primitives and resources

Availability of third-party integrations and tools

GitOps Support

Integration with Git-based workflows

Support for declarative configurations

Multi-cloud and Hybrid Cloud Support

Ability to work across different cloud environments

Support for on-premises and edge deployments

Package Hydration

Templating and parameterization capabilities

Ability to create and manage reusable, customizable packages (specialization)

Automation and CI/CD Integration

Support for automated deployments and updates

Integration with existing CI/CD pipelines

Observability and Monitoring

Built-in monitoring capabilities

Integration with popular monitoring and logging tools

Compliance and Security

Support for security policies and compliance standards

Secret management capabilities

Extensibility

Ability to create custom resources and controllers

Plugin architecture for extending functionality

Rollback and Version Control

Support for easy rollbacks

Version control of configurations and deployments

61 of 152

Technology Deep Dive

  1. Ansible and AWX (Tal Liron, Google)
  2. TKO and Porch (Liam Fallon, Ericsson)
  3. Kubenet’s Choreo (Wim Henderickx, Nokia)
  4. Porch Evolution (István Kispál, Nokia)
  5. Beyond Config Sync (Gergely Nagy, Ericsson)

62 of 152

What Is Ansible?

  • Infrastructure as Code (IaC)
    • Infrastructure = provisioning, deployment, configuration, etc.
    • Code = YAML + Python scripts (“playbooks”)
      • Can run remotely (ssh, etc.) or locally
  • Very mature
    • Since 2012 (12 years!)
    • Active support and development (acquired by Red Hat in 2015)
  • Very agnostic
    • Many platforms and technologies
    • Large ecosystem of ready-to-use plugins on Ansible Galaxy
    • Access to full Python ecosystem
    • GPLv3
  • Very straightforward
    • Data model = tasks and inventories
    • Not an engineer? Use YAML!
    • Engineer? Drop down to Python! (and Python can wrap anything)

63 of 152

What Is Ansible AWX?

  • Managed Ansible (used to be called “Ansible Tower”)
  • Farms out Ansible playbooks within a cluster
  • Cluster = custom or Kubernetes, via an operator (as seen in this demo)
    • Distributed scaling
  • PostgreSQL data backend
  • REST API + CLI + web GUI
  • Beyond playbooks:
    • Containerized playbook execution environments (custom)
    • Inventory integration and management
    • Pull playbooks, roles, and modules from Git
    • Simple (yet powerful) workflows
    • Schedules, notifications, approvals
    • Permissions (organization, teams, users, projects)

64 of 152

What AWX Can Do for Nephio

  • AWX can be Nephio’s orchestrator
  • We continue to use our KRM package data model
    • No change to our blueprints and user stories
  • Nephio code = collection of Ansible plugins (roles and modules)
    • We can develop directly in Python
    • Or wrap Go (or other) code in Python
    • Leverage Ansible and Python (and other) ecosystems
  • Playbooks to do the work
    • Provisioning Kubernetes workload clusters on baremetal or clouds
    • Configuring the infrastructure (hardware and networking)
    • Network function hydration:
      • Fan-out blueprints to inventory of sites�(In this demo we use TKO’s data backend. Can we use Kubenet’s pkgserver?)
      • Specialization�(Can we use Kubenet’s Choreo?)
      • Hand-off to deployment mechanism�(e.g. Config Sync and others)

65 of 152

AWX and Existing Orchestration Narratives

  • Ansible and AWX are already widely used
    • Greenfield (Nephio) and brownfield under a single umbrella
    • Hybrid cloud scenarios
  • A single playbook can trivially employ many technologies
    • OpenStack, Kubernetes, Helm, TOSCA, NETCONF/YANG, git, ssh, BMC, BPMN, etc.
  • Combine several playbooks into a workflow
    • Also: workflows of workflows
  • AWX can sit under or side-by-side with other orchestrators (e.g. O-RAN)
    • They can call AWX’s REST API (or the CLI) (as seen in this demo)
    • Of course, Ansible playbooks can call other orchestrators’ APIs
    • And you can even use Ansible playbooks to orchestrator AWX (as seen in this demo)
  • Sync existing site inventories into AWX inventories
    • Write inventory sources in Python (as seen in this demo)

66 of 152

Live Demo

  • Check it out yourself!
    • Install TKO (can run in a Vagrant VM)
    • Continue with Ansible AWX demo guide

67 of 152

TKO and Porch

  • Installed and tested TKO over a period of 3 weeks
  • Experimented with sites, templates and plugins and a little with deployments
  • Deployed multiple sites in kind using site template and the TKO kind plugin (slightly modified)
  • Used a simple template to deploy nginx on a site created on kind
  • Amended the more complex template that’s available in TKO, it partially worked (some deployments failed due to environmental constraints)

68 of 152

TKO and Porch: Learnings

  • It is much more straightforward to work with a DB backend than with git/gitea as the backend
  • The TKO plugin concept of having an API on which one could write adaptation plugins is powerful:
    • Python/kpt plugins supported in TKO
    • Other plugins can be added (stock plugins from Nephio or vendor or operator supplied plugins)
  • TKO is a PoC
    • more work is required to systemize TKO (the “ities”)
    • Some of the mechanisms are a little difficult to follow, such as topology templates
    • The architectural principles of TKO are quite different to the current Nephio/Porch architecture

69 of 152

TKO Inspiration in Porch

  • Porch is the current package handling system in Nephio
  • Porch can be evolved and improved with inspiration from TKO
  • Use DB backend repositories in Porch
    • Much more straightforward for package manipulation than using a git/gitea backend, especially for automatic/simultaneous/asynchronous changes to packages
    • Use git repositories for upstream and downstream
  • Use plugins for package transformations rather than Porch Tasks
    • The decorator architectural pattern
    • Package transformations can be customized and chained
    • Support for package transformation plugins (kpt functions, Python etc)

70 of 152

Porch Database Live Demo

  • A Porch backend that uses Postgres
    • Check out this Porch PR
    • Continue with the Porch DB Backend guide

71 of 152

Choreo: How did we come here?

KRM -> API -> Extendable

Reconcilers -> Day-2 operations

Etcd

Orchestration/�Specialization

Change Management

Version�Control

Packages

72 of 152

Choreo

KRM -> API -> Extendable

Reconcilers

Version�Control

choreo

73 of 152

Choreo

KRM -> API -> Extendable

Reconcilers

Version�Control

removed all the core k8s API(s)

choreo

simplified the reconcilers (starlark, go templating, jinja2) + added run to completion

support dev and prod (live)

support change mgmt

operates client/server side�enables collaboration

state

config

74 of 152

Choreo demo

75 of 152

Choreo

- Simplification:

- starlark, go templates, jinja2 templates, etc

- linux philosophy (do 1 thing and 1 thing well)

- Dependency management: a lower level abstract impact to higher level abstractions

- Hierarchical choreo resources (upstream ref)

- Dev and Prod (state and config)

- Client side/server side

- Collaboration

- Scale:

- Server side apply (granularity at the parameter level)

- Event driven

- Predicates

- Reconciler Goroutines

- Changes are only push when needed

- Backward compatibility

76 of 152

Porch Evolution

My answer to what’s Nephio’s unique selling point is…

Hydration

or more specifically the user experience it provides

Intents can be so complex that they are �created by multiple actors through multiple steps.

77 of 152

Porch Evolution - Why Nephio?

  • Declarative Intents + Reconciliation loops → proven method to decompose complex automation problems to smaller maintainable control loops
  • Coexistence of imperative configuration changes and declarative intents → Support out-of-band changes by humans and lets them co-exists with closed automation loops
  • Interoperability between controllers coming from different vendors
  • Single source of truth for all kinds of configurations (infra, LCM, app/NF config, complex high-level intents)
  • Version Controlled - Audit log, return to previous state
  • Direct access to low-level abstractions (KRM), instead of defining ad-hoc DSLs
  • Should be scalable - Programmatically generate and actuate 100s of similar configurations

78 of 152

Porch Evolution - same old Nephio R1

79 of 152

Hydration process with PackageVariants (Isti)

Staging�Package

Package�Variant

Staging�Package

Deployment Package

Deployment Package

Deployment Package

Package�Variant

Package�Variant

Package�Variant

Package�Variant�Set

fan-out

Blueprint

Package

Package�Variant

package �vendor’s�domain

customer’s�domain

Approvals and manual editing of configuration makes sense

fully automated closed loops, �no approvals, no manual edits

Published PackageRevisions �of the package

draft

The draft PackageRevision only exists temporarily here, until specialization produces a “ready” state. Then it is automatically approved/published

Staging�Package

Package�Variant

Staging�Package

Package�Variant

Staging�Package

Package�Variant

Merged Data

80 of 152

Porch Evolution

81 of 152

Porch Evolution - embracing readiness gates

Upstream

Package Revision

Package Variant

  1. Clone + set readiness gates (atomic)
  2. Apply mutations
    1. Built-in mutations (implemented in PackageVariant Controller)
    2. Custom mutations (specializers implemented in external controllers)
    3. Manual mutations
  3. Validate
  4. Readiness check
  5. auto propose, approve

Downstream

Package Revision

82 of 152

Porch Evolution - built-in mutations

  • Currently supported mutations:
    • Inject KRM object from the cluster (legacy config injection improved)
    • Inject a KRM object from a package other than the upstream
    • Inject a whole subpackage
    • Inject object from a manifest inlined into the Package Variant’s spec
    • Manipulate the pipeline of the package
  • In progress and planned mutations:
    • Merge external package to upstream
    • Inject inventory item
    • Run Starlark function
    • Run a KRM function
    • … run an Ansible playbook
    • … run Choreo to completion
    • … whatever the community decides to be useful…

83 of 152

Porch Evolution - mutations

  • Mutation list is very similar to an Ansible playbook/TKO functions/workflows/…
  • Package Variant mutations vs. kpt package pipeline
    • PackageVariant mutations are meant to be defined by “deployment engineer” role. They are applied exactly once after cloning.
    • The primary owner of the Kpt package’s pipeline is meant to be the blueprint package vendor. Pipeline is run after each and every change in the package.
  • Injection-type mutations enable value backpropagation from the workloads in the clusters to git
  • Built-in mutations are meant to be generally useful in a wide-range of scenarios
  • Custom mutations are usually specific to one kind of package

84 of 152

Beyond Config Sync

Feature comparison of gitops reconcilers and possible avenues of continuation

85 of 152

Gitops reconciler for O-RAN NFO

  • 100k Edge sites in a RAN setup
  • Extremely resource constrained edge sites
  • Overall footprint of gitops reconciler is huge if running in distributed mode
  • Upgrades need to happen in a short time window, can’t wait for periodic reconciliation

86 of 152

Config Sync

  • Can operate in distributed setup (the Anthos version can somehow integrate with GCP to operate in )
  • Doesn’t assume the existence of a management cluster
  • Proposes a more “gitops-purist” approach, but doesn’t offer a solution for the bootstrapping of the gitops reconciler itself.
  • Doesn’t support git subscriptions

87 of 152

FluxCD

  • Can operate in both distributed and centralized or federated setups
  • Doesn’t assume the existence of a management cluster
  • Proposes a more “gitops-purist” approach, all the config exists in git, it’s reconciled by walking the reconciliation tree
  • Has first-class helm support

88 of 152

ArgoCD

  • Can operate in centralized or federated setups
  • Assumes the existence of a management cluster
  • Proposes a blueprint definition in git deployment definition in management cluster approach
  • Contains some limited inventory-like interfaces like Application and ApplicationSet

89 of 152

Drop-in replacement for ConfigSync

FluxCD

ArgoCD

Can use CAPI’s kubeconfig format (base64 kubeconfig in a secret)

Yes

No

Can use Token reconciler’s format of secrets (username, password, token keys in a secret)

Yes

Yes

Can filter out config.kubernetes.io/local-config resources

Yes

No

90 of 152

Comparison

ConfigSync

FluxCD

ArgoCD

Distributed Setup

Yes

Yes

No (can be federated)

Centralized Setup

No

Yes

Yes

Git Webhook notifications

No

Yes

Yes

Documentation

Sparse

Extensive

Extensive

91 of 152

Project resources

Create an Account:

Follow LF Documentation at: https://docs.linuxfoundation.org/lfx/sso/create-an-account

Please subscribe to the Nephio developers mailing list. You can do that by sending an email to this address: nephio-dev+subscribe@lists.nephio.org You will receive an auto reply requesting subscription validation. The email content is not important.

To join as Supporter:

There are no documents to sign or fees to join, just the form needs to be filled out with the requested information. https://nephio.org/contact/

92 of 152

93 of 152

Nephio R4 Developer Summit - 2024 (Day 2)

94 of 152

Lightning Talk - Service Experience and GenAI (Tomas Achaval - Wilab)

95 of 152

Lightning Talk - Paraglider (Sarah McClure)

96 of 152

Streamlining Cloud Networking

97 of 152

Agenda

  1. Motivation
  2. Approach
  3. Interface
  4. Architecture
  5. Roadmap
  6. Potential Collaborations
  7. How to Get Involved

98 of 152

Setting: Enterprise Cloud Deployments

  • Cloud users with networks connecting resources of various types
    • VMs, clusters, managed services, etc.

VM

Storage Service

Kubernetes Cluster

?

99 of 152

Motivation

  • Cloud Networking is Complex
    • Abstractions often are virtual versions of physical components
      • Virtual networks
      • Subnets
      • Security rules
      • Gateways
      • Peerings

100 of 152

Motivation

  • Cloud Networking is Complex
    • Abstractions often are virtual versions of physical components
    • Have to build the network necessary to achieve high-level goals
    • Different versions of each component in each cloud
    • Multicloud networks require more components and coordination

101 of 152

Approach

Elevate the abstraction by focusing on connectivity needs between resources

Paraglider provisions the network necessary to achieve the requested connectivity

102 of 152

Approach

  • Connectivity expressed as permit lists on resources
  • Rules can reference names or groups of resources
  • Specify intent for network functions like load balancing on endpoints/connections

Cloud network configuration

103 of 152

Interface

  • Modify permit lists
    • Ex: glide rule add gcp resource_name <rule>
  • Name / Group resources with tags
    • Ex: glide tag set frontend_vms —children vm1,vm2
  • Create / Attach resources
    • Ex: glide create resource azure vm_name <vm_config>

104 of 152

Architecture

Orchestrator

Azure Plugin

GCP Plugin

IBM Plugin

Tag Service

User Requests

Paraglider Controller

105 of 152

Example:

Goal

Azure

GCP

A

(us-west)

C

B

(us-east)

106 of 152

Example: Paraglider Steps

  • Start Paraglider services locally:��glided startup <path_to_paraglider_config>
  • Create a VM with Paraglider enabled on Azure and GCP:��glide resource create <azure|gcp> <resource_name> <path_to_vm_config>
  • Add rule to enable access to VM:��glide rule add <azure|gcp> <vm> --<ssh|ping|…> <ip_range|tag|other_vm>

107 of 152

Example: Paraglider Steps

  • Create a VM with Paraglider enabled on Azure and GCP:��glide resource create azure vm-a <path_to_vm_config>

glide resource create azure vm-b <path_to_vm_config>

glide resource create gcp vm-c <path_to_vm_config>

  • Add rule to enable access to VM:��glide rule add azure vm-a ping default.azure.vm-b

glide rule add azure vm-b ping default.azure.vm-a

glide rule add azure vm-a ping default.gcp.vm-c

glide rule add gcp vm-c ping default.azure.vm-a

108 of 152

Example:

Under

the Hood

Azure

GCP

A

(us-west)

C

B

(us-east)

109 of 152

Other Features

  • Supports Kubernetes clusters and private endpoints for services, as well as VMs/instances
  • Supports connections to public endpoints
  • As tag membership changes, permit lists automatically update
  • Namespaces provide a unit of isolation between deployments
    • Different infrastructure, but can create connections between them

110 of 152

Roadmap

Ongoing Work

  • Brownfield deployment compatibility
  • Cloud service support
  • AWS plugin

Longer Term

  • Network function support
  • Support for more clouds
  • Tight Kubernetes integration
  • Terraform support
  • Eventual goal: first-party support from cloud providers

111 of 152

Opportunities for Collaboration

  • Provision the underlying (networking) infrastructure for Nephio
  • Provision infrastructure necessary for multicloud deployments
  • Design collaboration for tighter Kubernetes integration in Paraglider

We’re excited to hear from the Nephio community about other ideas!

112 of 152

How to Get Involved

  • Join our Discord
  • Check out our issues on GitHub
  • Use Paraglider

All the above can be found at paragliderproject.io

113 of 152

Lightning Talk & Demo - Nephio for GenAI Automation Use cases (Josh Joseph, Stanford)

114 of 152

Lightning Talk & Demo - Nephio for GenAI Automation Use cases (Josh Joseph, Stanford)

115 of 152

Nephio’s External APIs

Ciaran Johnston, Ericsson

116 of 152

What APIs do we expose?

  • Custom Resource Definitions
    • Package Orchestration
    • Topology and Networking
  • O-RAN APIs
    • FOCOM
    • NFO
    • O2IMS
  • Git / OCI
    • Formal contract?

117 of 152

Custom Resource Definitions

Package Orchestration

https://docs.nephio.org/docs/apis/porch/

  • This is not up to date for some reason – git repo here

Number of APIs have been reduced from 14 to 9

All existing APIs are at “v1alpha1”

  • Stability would entail getting to “v1” (via v1beta1)
  • Need to determine “MVP” level of functionality

Most important elements:

  • API Extensions: PackageRevision, PackageRevisionResources (PackageRev ??)
  • Custom Resource Definitions: Repository, PackageVariant, PackageVariantSet

A number of elements have been removed / cleaned up already from the API

118 of 152

Porch API Stabilisation

Phase 1: stabilize fundamental APIs:

  • Repository – everything starts here
  • PackageRevision, PackageRevisionResources – fundamental Porch components

Phase 2: stabilize higher-level APIs:

  • PackageVariant, PackageVariantSet – less well defined, builds on the other APIs

119 of 152

Porch API Stabilization

Repository

  • Remove “upstream”, “mutators” and “validators” from current spec until defined how they will be used
  • Leave OCI as an alternative backend - more types may be possible?
    • Support needs to be completed to stabilize the API - worth it?
    • Need to document how to add other backend types - filesystem, DB etc?

GitRepository struct

  • Need to document how secrets are handled including mTLS (unclear if there are API implications here)

120 of 152

PackageRevision

PackageRevision specification

  • “WorkspaceName” attribute is used to allow for multiple simultaneous drafts - could this be handled by having multiple “PackageRevision” instances in draft stage?
    • Should we avoid multiple drafts until we have a stable approach?
  • ParentReference may not be used? This is not a reference to git commits, since multiple packages exist in the same git repo
  • Drafts are all destroyed once approved, so ParentReferences are to previous approved package revisions
  • Are we just reinventing git?

PackageRevision 1

PackageRevision 2x

PackageRevision 2y

PackageRevision 2

PackageRevision 3x

PackageRevision 3

Implicit parent relationship

Implicit parent relationship

Forced rebase

121 of 152

PackageRevision #2

Tasks

  • Audit trail, instructions for updates, and patches. Exposes kpt imperative commands.
  • Complicated and does not always work, not extensible
  • Should it be present as part of the external API?

Upgrade usecases:�1. Upgrade with new blueprint (inherit what you can from previous PR, add in stuff from blueprint, and add/modify instance specific configuration) -- If this doesn't need to be supported, then tasklist can be discarded.

2. Upgrade with new blueprint without inheriting (add stuff from blueprint, add instance specific configuration)

3. Upgrade/update without blueprint change (inherit from previous PR, and add/modify instance specific configuration)

Is there a better way than tasklist?

122 of 152

O-RAN APIs in Nephio

OAI- CU-C/U

OSC INF Edge Cloud (Nephio Workload Cluster)

SMO Services

RAN Domain Management

O-Cloud

SMOS

Communication

A1

IMS

O1

O2ims

DMS

E2

E2E Services

Service Management and Exposure

Data Management and Exposure

End to End Service Orchestration and Slice Management

Service

Orchestration

Service

Assurance

RAN

Analytics

AI/ML

Workflows

Software Packaging Onboarding

Non-RT RIC <rApp>

RAN NF OAM

FM|CM|PM

Physical Network

RU

O2dms (K8s profile)

OFH

E2

OSC SMO (Nephio

Management Cluster)

OSC INF Central Cloud (Nephio Management Cluster)

OSC Near-RT RIC <xApp>

monitori

inventory

provision

lcm

O1

O1

OAI-DU�-RF-SIM RU-

K8s Operators

Topology and Inventory

NFO

Nephio NBI Adapter

FOCOM

Nephio SBI Adapter

Nephio NBI Adapter

OSC Software (Will discuss)

Target CNF for deployment

123 of 152

O-RAN APIs in Nephio

Nephio should play a role in O-RAN as three different components

  • FOCOM - request clusters
  • NFO - request workloads
  • IMS - manage clusters for running workloads *

Supporting these roles requires implementing multi-vendor interfaces and supporting Nephio in 3 different roles interworking with each other

  • And potentially with 3rd party implementations of the same interfaces

Nephio modelling will refer to O-RAN APIs and models for these interface specifications

* IMS overlaps with Sylva?

OAI- CU-C/U

OSC INF Edge Cloud (Nephio Workload Cluster)

SMOS

Communication

IMS

O2ims

DMS

Service

Orchestration

Service

Assurance

O2dms (K8s profile)

OSC SMO (Nephio

Management Cluster)

OSC INF Central Cloud (Nephio Management Cluster)

monitori

inventory

provision

lcm

OAI-DU�-RF-SIM RU-

K8s Operators

NFO

NFO NBI Adapter

FOCOM

IMS SBI Adapter

FOCOM NBI Adapter

OSC Software (Will discuss)

Target workloads

Multivendor integration points

IMS NBI Adapter

124 of 152

Git / OCI as a formal interface

  • Porch provides a Kubernetes API for manipulating packages, their contents, and how they map to repositories (which may be Git or OCI repositories)
  • Some operators want to use git code review processes as the human way of interacting with package contents
  • CI systems may onboard packages / blueprints into an SMO using OCI registries

Does the Nephio community agree with these means of integration towards Nephio?

125 of 152

WG6: GenAI and Nephio

Sana Tariq, Telus

126 of 152

What GenAI Means for Nephio

Templates generation

GenAI based Operators, CRDs, TOSCA, KRM etc.

Templates hydration

GenAI based data-fill for various environments and context

Nephio User Automation Simplification

Nephio Services Closed Loop

SDKs and APIs

GenAI based SDKs and APIs creation

Cloud Optimization

AI optimizing cloud capacity energy and cost

5GC, RAN and Edge

AI optimizing performance, efficiency and reliability

Network Operations

GenAI provisioning and troubleshooting agents

127 of 152

Nephio and GitOps

Custom Resource

Operator (Reconcile)

Current State

Notification

Change Events

Tracking

Act

I want an eMBB network slice with SLA minimum download speeds of 100 Mbps and latency below 20 millisecond

Deploy 5G core network with a capacity of supporting 10 million concurrent connections and integrate with an ORAN…

Deploy 5G core CNFs SMF, AMF, UPF and Configure ORAN vDU/vCU with xyz…

Intent

Observability tools, notifies about drift in actual state and desired state

Humans writing CRDs and verifying intent

Ensures the running infrasture converge to the desired state notified and declared in CRDs

128 of 152

WG6: Setting up LLMs RAG and next steps…

Large Dataset

Retrieval Model

Pre-trained LLM

Prompt

Response

Prompt

Prompt

+

Code Repo

Config Files

Network troubleshooting and performance analytics

Search

Relevant Documents

Nephio/Organization/ Domain specific Data

Response

129 of 152

WG6: Focus areas and Next Steps

Nephio R4

  • WG6 Meetings

Friday 9:00 am PT

  • Continue progress on White Paper (GenAI and Nephio)
  • Identify use-cases to focus on
  • Setting up LLMs RAG pipeline
    • Plans review
    • Budget review
  • Identify R5+ use-cases

Nephio R5

  • Cards, Operators and SDK generation with RAG pipeline
  • Produce candidate user-stories with other WGs
    • WG2: O-RAN
    • WG5: Service Assurance

Nephio R4 and Beyond

  • Community Growth
    • Introduction of AI startups
  • Adoption in Nephio Project
    • Template Generation
    • Templates Hydration through integration with Nephio Core Tooling
  • Enabling AI capabilities in Nephio SDKs, UI and Core technology

130 of 152

Call for Community!

Newly created Working Group (SIG1/WG6) especially for GenAI Use Case!

Welcome Vendors, Providers, Startups!

131 of 152

Nephio progress Testing/Release

Rado, Bala

132 of 152

Agenda

  1. Current status
  2. What’s in works for R4
  3. Plans for the future ahead

SIG Release

133 of 152

Current architecture status

SIG Release

GitHub

Prow

Code submission

GitHub Interaction

Status

Reports

Webhook Events

Job Results

Dashboard

One place to check the results and trends

GitHub Bot

Prow ‘slash’ commands

Keeps repository policies applied

Keeps user roles defined and abide

Runs jobs on code/code changes

Simplifies handling code review and merging

GitHub

Actions

Runs some artifact builds and code checks

Docker Images

VM for

E2E tests

134 of 152

What if?

SIG Release

GitHub

Prow

Code submission

GitHub Interaction

Status

Reports

Webhook Events

Job Results

Dashboard

One place to check the results and trends

GitHub Bot

Prow ‘slash’ commands

Keeps repository policies applied

Keeps user roles defined and abide

Runs jobs on code/code changes

Simplifies handling code review and merging

GitHub

Actions

Runs some artifact builds and code checks

Docker Images

VM for

E2E tests

135 of 152

What’s in works for R4

SIG Release

Move some jobs to GitHub Actions

  • “Costless”
  • Runners maintenance outsourced
  • Actions developed externally
  • On downside: no dashboard, Prow knows nothing about them

Run simple E2E test on PR content

  • Lower the time to debug eventual issues
  • Keeps an eye on standards
  • On downside: takes time and has some complexity into it

Release process automation

  • Process resilience to staff changes
  • Takes less time
  • Repeatability
  • SLSA requirements

Development process standardization

  • Defines who’s who
  • Defines responsibilities and privileges
  • More welcoming environment for newcomers

136 of 152

Plans for future ahead

SIG Release

  • More automation
  • Seamless Prow - GH Actions integration
  • Or switching to only using GH Actions for everything
  • Full SLSA 4
  • Migrate to lighter base images

137 of 152

WG7: Nephio Transport Use Case

Evgeniy Zhukov - GlobalLogic

138 of 152

Agenda

  1. Why Transport is needed
  2. Transport User Story scope for R4
  3. Nephio Transport Roadmap

139 of 152

Why Transport is needed

  • Long distance communication
  • Traffic aggregation

140 of 152

Transport User Story scope for R4

Deploy 4 Network Elements:

  • Core
  • Edge01
  • Edge02
  • Edge03

MPLS Tunnels:

  • N2
  • N3
  • N4

Automation:

  • Nephio-Kubenet Integration
  • 5GS dictating Transport configuration

User Story:

AUSF

NRF

NSSF

SMF

gNodeB

UPF

AMF

Regional Cluster

Edge Cluster

Edge Cluster

Edge02

Edge01

Edge03

Core

N4

Tunnel

N2

Tunnel

N3

Tunnel

Transport Cluster

141 of 152

Nephio Transport Roadmap

R5

    • Basic
      • Nephio-Kubenet integration
      • Automation

R6

    • HA
      • Carrier Grade Availability >99.95%
      • Loop topology with redundancy protocol

R7

    • DC
      • Data Center Leaf and Spine Architecture
      • Data Center + Transport + 5GS Networking

142 of 152

Call for Community!

Newly created Working Group (SIG1/WG7) especially for Transport Use Case!

Welcome Vendors, Providers, Startups!

143 of 152

SIG Security Workload Identity -Rahul

144 of 152

SIG4 Security Charter Progress

We started with a score of 3.6

145 of 152

Workload/Service Identity

  • Top Use-Cases?
  • Why SPIFFE?
  • Current Progress?
  • R4 Plan?
  • Immediate next Steps?

146 of 152

Workload/Service Identity Provisioning

147 of 152

R4 Plan

  • Workload/Service Identity PR
  • Nephio Infrastructure as Code scanning
  • OpenSSF best practices improvement scope?
  • Improving Security Response

148 of 152

R4 Planning and discussions

Bala, Tal, Sana, Rahul, Wim

Review R4 user-stories

Identify ownership

Finalize R4 scope

149 of 152

SIG1 Architecture [ Sana ]

149

150 of 152

SIG2 Automation [Tal & Wim]

  • Planning work for O-RAN user stories for R4
    1. Deploy O-Cloud Kubernetes cluster
    2. Deploy NF using SMO
    3. Terminate NF using SMO
  • Various fixes and improvements to Porch and Nephio reference implementation

150

151 of 152

152 of 152

Realizing Nephio Business Value for Telcos

Daniel Freiji

Director Telco Cloud

Telus Canada

Daniel Bernier

Technical Director, Bell Canada

Bernard Tsai

Lead Architect

Deutsche Telekom

Germany

David Kinsey

Lead Principal Architect, ATT, USA

Moderator

Sana Tariq Telus

Eric Debau

Head of R&D

Orange, France