Modernize the Network

Modernize the Network

迎接5G 、邊際運算時代來臨，連接未來轉型關鍵

Eric Leung – Director, DellEMC

2 of Y Dell – Internal Use – Confidential

Trends driving the 5G visionUse-Cases, Competitive Pressures, Demand Driving New Imperatives

MARKET / TECHNOLOGY TRENDS INVESTMENT IMPLICATIONS

Content / Services(AR/VR, Gaming)

Edge/Distributed clouds

High-bandwidth, low-latency architectures

Internet of Things(Self-Driving Cars, etc.)

Telco Cloud(NFV, SDN, SD-WAN)

Need to be agile and offer new services

IT Workload migration to Cloud

Enterprise Mobility

Margin pressure,New Competitors

Drive 5G connectivity

Decentralization of infrastructure Upgrade and expand network

Automate virtualized network infrastructure


5G services uses cases

1

2

3


5G enhanced experiences

Rich, real-time user-generated content, like video sharing

Instant access to media and entertainment

Connected cloudcomputing

Low latency online multiplayer gaming

The basics: responsiveweb browsing

Interactive venue experiences like live 360o streaming

Mobile immersiveexperiences

Boundless photorealistic extended reality


Major 5G industry wide investment points

USERS1,000,000’s

AGGREGATION & METRO100’s

CORE10’s

ACCESS & NETWORK EDGE10,000’s

OPS1-2

NETWORK CORERADIO | ACCESS

FAR EDGE & EDGE

• 5G Enabled Devices• Consumer Electronics • Laptops /

Smartphones• IoT Gateways• Indoor Access / Small

Cells• Access (uCPE, SD-

WAN)

• 5G Spectrum• 5G NR NSA / SA• Small Cells /

Densification• 4G LTE Marco Cell• Deep Fiber

Investments• Microwave / Fiber BH

• Fronthaul Switching / CSR• C-RAN and vRAN• MEC Platforms w/ VNFs• Edge NFV-I / vEPC / NGC• Edge Ready Facility (MDC)

METRO TRANSPORT• IP Everywhere• Carrier SDN• MPLS / CE /

DWDM

• NFV-I Platforms w/ vEPC / vIMS / SGi-LAN

• Evolved to 5G Core (NGC)

• SDN Controllers• Open Networking SERVICE OPERATIONS

• MANO• BSS Modernization• OSS Modernization• Analytics & Data Lakes• Infra Monitoring, AI, ML• IT Infrastructure

SUBSCRIBER ECOSYSTEMS


5G System Architecture

5G System architecture - Reference point representation

§ 5G core network is now based “Service-Based Architecture” (SBA).

§ Centered on services that can register themselves and subscribe to other services.

§ 5G core platform will be more programmable and allow many different functions to be built, configured, connected and deployed at needed scale.

Ø Leveraging cloud software technologies, 3GPP NF architecture leads to higher flexibility, programmability, automation and significant cost/energy reduction.

Source: https://portal.3gpp.org/desktopmodules/Specifications/SpecificationDetails.aspx?specificationId=3144


5G Edge and multi-cloud

• Distributed solutions• 5G optimized for core, edge and

far edge NFV-I workloads• 5G NR and NGC performance

targets, capacity needs, cloud native and CUPS driven architectures demand an evolved NFV-I platform

5G VNFs & APPLICATIONS

DISTRIBUTED CLOUD

SMARTEDGE

VM &CONTAINERWORKLOAD

AGNOSTIC ACCESS

Virtual Storage

Virtual Network

Virtual Compute VIM

FPGA | GPUSSMARTNICS

SELF OPTIMIZINGNFV-I

+

CONTROLPLANE

DATAPLANE


5G and AI

5G NG Core brings cloud centric capabilities into a highly virtualized and increasingly automated framework

Dynamic workload placement driven by intelligent network slicing use cases

E2E 5G SYSTEMS

REAL TIME TELEMETRY

DATA LAKES

AnalyticsAI & ML

Resource Management

Service Management

Slice A

Slice B

Cloud Native Stateless VNF

Control & User Plane Separation

Open APIs Data Model Distributed

Management

Containers VMMicroservices


5G Cloud Native ApplicationMoving towards from monolithic to microservices type of architecture

Monolithic VNF

• VNF combined with all the components.

Minimal Reuse

• VNF level reusability is low or even no

Stateful

• The states are tightly coupled with the packet, processing unit itself.

Orchestration

• VNF orchestration is depending in VNF complexity.

Containerized VNF

§ Reduced overhead

Microservices

§ Which enables service composability, reusability, efficient scaling and ease of deployment

Stateless Processing

§ Makes cloud VNFs fault tolerant and scalable without limit –Application isolated from state.

Easy Orchestration

§ VNFs are designed to minimize the amount of configuration needed in each component.

CN - VNF Current Application (VNF) Cloud Native Application (VNF)


Why Micro-services ?Compose to gain scalability, flexibility and reusability

Monolith Microservices

Service 1 Service 2 Service 3 Service 4

Presentation

Business

Persistence

Service 1 Service 2

Service 3 Service 4

Microservices architecture: § Components can be independently updated/delivered§ May use containers for each of the components§ Monolith vs componentized§ Each component can evolve independently


Distributed workloadsWorkload will be more distributed vs centralized workload


Containers Deployments Considerations


Kubernetes key concepts

• Cluster, Node, Master, POD, Container runtime• Label & Label selectors• Annotations• Replication Controllers & Replica Set• Services• Scheduler, Volume• Name/namespaces• Deployment , DaemonSets, Statesfull sets


K8s - Multiple Network Interface for VNFsPROBLEM

• In NFV use cases, one key requirement is the functionality to provide multiple network interfaces to the virtualized operating environment of the VNF.

• Kubernetes support only one Network interface –“eth0”

USE CASES

• Functional separation of control and data network planes

• Link aggregation for redundancy of the network• Support for implementation of different network

protocol stacks and/or SLAs• Network segregation and Security

REFERENCE

Multus CNI –https://github.com/Intel-Corp/multus-cni


K8s - VHOST USER CNI PLUGIN Integration with MULTUS


ONAP – Support for K8S based remote regions

Site(With Openstack VIM)

Site (*)(Two logical edges –

Openstack for VMs and K8S for containers)

Current§ Support of Openstack based remote Clouds, § Support multiple Openstack variations –

Windriver Titanium, VMWare VIO, Native Newton, Ocata. Only VM based VNFs.

Need§ Need support for containerized VNFs.§ Need support where K8S is used for both

VMs and Containers (To avoid multiple controls and also to utilize same compute nodes for both VMs and containers)

Source: ONAP Project

Site(With K8S for both VMs and

Containers)

ONAP

Multi Cloud Service

APP Orchestrators OSS/BSS

SDNC(Fabric/WAN Control)


ONAP – VM and Container networking (Uniform)

ONAP

Multi Cloud Service

APP Orchestrators OSS/BSS

SDNC(Fabric/WAN Control)

Compute Servers – Hardware (CPU/Memory/SRIOV-NIC/FPGA/GPU) etc…

NFVI/VMM

Network1 (VLAN/VXLAN/GENEVE)

Network N (VLAN/VXLAN/GENEVE)

SRIOV networkingvSwitch (e.g OVSDPDK) networking

VM VMC C

POD

CPODSite (edge/cloud)

K8S Master

Network Controller

Current: Networks are created and VM are placed in various networks (VLs and CPs – Virtual Links and Connection Points).

Need: Some workloads can be containers and networking is expected to be extended to containers

Source: ONAP Project


5G Service Characteristics

High Speed High Volume

Lower Latency

Personalized

Close to the Consumer

Reliable

Immersive

On Demand

Flexible / Scalable

Agile

Consistently Available

Lower TCO

Modernize the Network

Documents