WIXLIB

Kubernetes Native Infrastructure andOperator Framework for 5G Edge Cloud ComputingHyde Sugiyama, Chief Architect Red Hat K.K.

OpenShift/K8s on OpenStack NFV session@OSS2018OpenShift on OpenStack NFVB2B2X for SoE apps in OpenShiftVM) on top of DCN(NFVI)2(in

NFV Evolution to er30％70％(Stateless & Stateful）(IT SoE apps）VNFLong re MetalBare MetalBare MetalData LakeData LakeData LakeSoR: System of RecordSoE: System of EngagementOCP4: OpenShift Container Platform 4

Agenda 4Kubernetes on bare metal deployment5GC workloadsO-RAN allianceEdge computing workloadsSummary

Innovation - 20194GAll IP packetCarrierGradeLinux MultiaccessEdgeComputingVNF55G5GCCloud native/Service Based ArchtectureCP and UPF separationNetwork slicingUPF offload (FPGA, Edge Switch Fabric)vRAN CU-DU splitHeterogeneous ComputingDPDK(vCPU)GPUFPGACNFAutonomous micro edge cloudEdge AI platform(Intelligent Edge)Data Lake /Data Hub

OpenShift/k8s on BMdeployment6

Performance Sensitive Application PlatformKubernetes can enhance (withadapting from more significantopen sources) like an OpenShift 4that becomes the single platform5G/EdgeMachineLearningFSIHPCISVsto run any application. 7Old or newMonolithic/MicroserviceBig DataAnimation

Possibility for NFV architecture change by adaptingKubernetes Operator(w/ OpenShift Operator framework) VM VNF guestOS VM VNF guestOS VM VNF guestOS Hypervisor NFVI (host OS)8API managementNFVO/OSS/BSSNFVOOSS/BSSVNF-M Pod Pod Pod CNFCNFCNF(SMF)(UDM)(.)CNF-MCNF-MCNF-MCluster OperatorVIMOpenShift Worker(host OS)BM(s)BM(s)VNFCNFOpenShiftMaster

Kubernetes Operator Automate day 2 lifecycle management ofcontainerized applications in KubernetesLeverage CRDs to deploy Kubernetes nativeservices that can access Kube API eventsOperator SDK simplifies creation of Operators inGo (or leverage Helm or Ansible automation)Helm Operator allows you to convert Helm Chartsinto Operators Deploy Charts without requiring Tiller Leverage Kube RBAC to deploy Charts Automated, over the air updates for Chartdeployed on pattern of "SRE (Site Reliability Engineering)"that incorporates operations by softwarehttps://www.youtube.com/watch?v LymzLHRbQdk9

Full-Stack Automated Operations in OpenShiftAPPLICATIONS AND SERVICESISV OperatorsCustom Operators (built w/Operator SDK)PLATFORM AND CLUSTER MANAGEMENTAutomated updates for Kubernetes,monitoring, security, registry and moreLINUX HOSTOver-the-air updates for Red Hat CoreOSANY INFRASTRUCTURE10

5GC Service BasedArchitecture on Kubernetes11

5GC system workData ionFunctionAccess &Mobility tSessionManagementFunction

KIALI ( GUI for Istio )5G SBI and Service MeshOPENSHIFT SERVICE MESH A dedicated network for service toservice communications13 Observability and distributed tracing Policy-driven security Routing rules & chaos engineering Powerful visualization & monitoring Will be available via OperatorHub

5GC SBA with Service MeshKNI(Kubernetes Native Infrastructure) for EdgeIstioservicemesh5GCMULTUS CNI setup,Stateful management,CNF LCM, K8s LCM,etcUser EquipmentKNI edgeGPU device plugin, DPDK, FPGA,Serverless, Data lake, unctionAccess &Mobility anagementSessionManagementFunction

5GC, UPF and Network Slicing (Cloud edge fabric case)Slice 1 forB2B2X workloads 1: SMF1 & UPF1Multi cluster for istioservice mesh might beneededSlice 2 forB2B2X workloads 2: SMF2 & UPF2**UE**15*https://www.youtube.com/watch?v ion/WhitePaper iPOP2019.pdf

O-RAN alliance &RAN Cloudification w/Kubernetes16

3GPP &O-RAN allianceDesignInventoryRAN Intelligent Controller (RIC)non-RTPolicy ConfigurationOrchestration & Automation (e.g. ONAP): MANO, NMSA13GPP (e/g)NBRAN Intelligent Controller (RIC) near-RTApplicationsE3rd party 2 Radio dModelE2 :btw RIC near-RT and O-CU/O-DUMulti-RATO-CU ProtocolStack3GPP DUMobilityMgmtRadio-Network Information Base3GPP CUOur -UPSDAPPDCP-UNFVI Platform: Virtualization layer and COTS platformO-DU: RLC/MAC/PHY-highOpen Front HaulO-RU: PHY-low/RFF1

O-RAN alliance WG6Decoupling of software from hardware for all RAN modules in all trationPolicy,MetricsBlackboxBBUO-DUO-CUH/W abstraction layerNFVI stack (containers or VMs), OS, cloud m-planeSupport 10,000sof distributedcloud ralized CU/DU(C-RAN)18All RAN modulesO-RUCU/DU splitDistributedCU/DU(D-RAN)Multitude of deploymentmodels: CloudRAN,CU-DU split,dRAN on whitebox or DC

Candidate WG6 Scenarios B,CScenario B: Distributed vO-CU and vO-DUBBU chassis at cell siteTo UPF& MECNear-RTRICNFVIvBBUvO-CU vO-DUE21:N, remoteRegional cloudNFVIOpen fronthaul1:m, localO-RUCell siteOpen chassis and blade specK8s or OpenStackCell site areaScenario C: Centralized vO-CU with distributed vO-DUThe O-CU/O-DU functionality can bepooled from multiple O-RUs, and meetO-DU latency requirements.Near-RT RIC can serve a very largenumber of O-RUsO-DU chassis at cell siteTo UPF& MECNear-RTRICvO-CUNFVIF11:N, remoteRegional cloudK8s or OpenStack19vO-DUNFVIOpen fronthaulO-RU1:m, localCell siteOpen chassis and blade specCell site areaThe O-DU functionality can be pooledfrom multiple O-RUs, and meet O-DUlatency requirements.Near-RT RIC can serve a very largenumber of O-RUs, and O-CU is verycentralized.

KNI EdgeKubernetesNativeInfrastructure20

Akraino Edge Stack projectKNI-Edge Blueprints (in Progress)Provider Access Edge (PAE)Industrial Edge (IE)Optimized for real-time andnetworking performance forContainerized vRAN and MEC workloads.Optimized for small footprint andlow-latency for IoT, serverless, andmachine learning workloads.vRAN(RIC)MEC Apps VM1 . VMNML AppsMEC NOSPrometheus exportersOpenShift/OKDCephCRI-OCoreOS and CentOS-rtSrv-lessML AppsIoT nShift/OKD OSCoreOS and CentOS-rt.switch COTSCOTS COTSswitch COTSCOTS -Native Infrastructure %28KNI%29 Blueprint FamilyUBI: -universal-base-image21

IoT Apps - Eclipse IoT projectCONNECTED“THINGS”IoTEDGEAPPLICATION DEVELOPMENT,DELIVERY, & INTEGRATIONIoTINTEGRATION nalyticsMachinelearningDATA MANAGEMENT & ANALYTICSTelemetryManagementMachine learning model22CLOUDERA’S DISTRIBUTIONINCLUDING HADOOP (CDH)

ML Apps - OPEN DATA HUBCollaborate on a Data & AI platform for the Edge Cloud & CoreA collection of open source andcloud components packaged in a“machine learning-as-a-service”platform to solve businessproblems.23

Collaborate on a Data & AI platform for the Edge Cloud & Core AI Library Data Science and ETLTools Streaming and EnrichingData24 Storing Data Managing Data Monitoring Infrastructure

Serverless Apps - Knative Familiar to Kubernetes users. Native. Scale to 0 and autoscale to N based on demand Applications and functions. Any containerworkload. Powerful eventing model with multiple eventsources. Operator available via OperatorHub Knative v0.6 (v1beta1 APIs) No vendor lock inLearn 25

Red Hat OpenShift Hybrid ServerlessDeveloper experienceAPIs, CLI, service bindingBuilding blocks for serverlessSource-centric andcontainer-basedThe leading enterpriseKubernetes platformAutomated OperationsBuild an run anywhere (HybridCloud)Function as a ServiceRed Hat OpenShiftODO se CheRH MW Services(Operator backed)ISV Services(Operator backed)Operator FrameworkAutomatedOperationsIstioHybrid Install / OpsInstall / UpgradeNetwork / CNIOps & Dev ConsolesSecurity / AuthStorage / CSIKubernetesRed Hat Enterprise Linux or Red Hat CoreOS26API

Knative Event SourcesUpstream Event Sources:GitHubKubernetes EventsPub/Sub (AWS SQS, Kafka, NATS, Google PubSub)WebsocketsExpose an ingress200 event sources through Camel-KCamel-K Knative Demohttps://github.com/knative/eventing-sources27

Serverless ：IoT ＆ vaScript(NodeJS)ArchivingEDGE28PrivateHybridPublic

Summary29

Next Generation DX platformApplication portability in any environment and consistency in operation.Application onOS dependenciesOSOS ndenciesPodPodPodPodPodEasy LCM per microservice, multi-cluster management and federationMulti Interface, SDN for eOSRHEL/CoreOSGuest VMVirtual MachineVirtual MachineVirtual MachineGPULAPTOPCode Ready ***30VIRTUALIZATIONRHV, VMWAREBy using UBI (Universal Base Image),developers can concentrate on containerapplication development at any platform.PRIVATE CLOUDFPGABARE METALPUBLIC CLOUDOpenStack NFVIntel, Power, ARM(plan） Azure, AWS, Google, etc.It is also possible to duplicate and store the samecontainer image in geographically distant places(utilization of Ceph data lake and QUEY)** https://github.com/code-ready/crc

Summary1. Heterogeneous Computing Platform2. K8s as AI platform at Telco Edgea.Data Lake/Hub at Telco node3. Kubernetes Native Infrastructure on Bare metal across TelcoCore and Telco EdgeOperator Framework for Site Reliability Engineering and Providerextensioni. Autonomous micro-cloud at Telco edgeii. Digital Service Provider driven CNF apps managementb. Service mesh for micro servicec. Serverless at Telco edge nodea.31

Thank you!32

Computing 5GC Cloud native/Service Based Archtecture CP and UPF separation Network slicing UPF offload (FPGA, Edge Switch Fabric) vRAN CU-DU split Network Function Carrier Virtualization Grade Linux DPDK(vCPU) GPU FPGA Heterogeneous Computing Data Lake /Data Hub Autonomous micro edge cloud Distribute Compute Node VNF CNF Edge AI platform .