Agenda ● Telco and 5G ● Network Functions Virtualization ● OPNFV and Software Defined Everything with SUSE ● Realizing NFVI requirements ● Performance ● Demo
Telco and 5G
LTE Network Architecture
Telco functions ● The infrastructure uses several specialized hardware equipment to realize network functions ● Radio Access Functions ● EnodeB ● Core Functions (Evolved Packet Core – EPC) ● MME - Mobility Management Entity ● SGW/PGW - Service/Packet Gateways ● IMS – IP Multimedia Service ● Edge functions ● Customer Premises Equipment – CPE ● Eg: Residential Gateway for Internet and VoIP and Set Top Box for Media
A feel for the Telco hardware EPC CPE SGW Outdoor EnodeB EnodeB
What are the problems ? ● Expensive ● Built to specifications ● Built by specific vendors ● Non-scalable ● Hard to maintain ● Difficult fault recovery
What does the specialized hardware do ? ● Legacy ● Specialized functions ● Reliability ● Consistently High Performance ● Security
The world of 5G
What is 5G ? As defined by ETSI: 5G is an end-to end ecosystem to enable a fully mobile and connected society. It empowers value creation towards customers and partners through existing and emerging use-cases delivered with consistent experience and enabled by sustainable business models. ● The realization of 5G is enabled by the following 4 components: ➢ NFV – Network Functions Virtualization ➢ MEC – Multi-access Edge Computing ➢ MWT – Millimeter Wave Transmission ➢ NGP – Next Generation Protocols
Network Functions Virtualization
What is NFV ? ● NFV reads as Network Functions Virtualization. ● VNF (Virtualized Network Function) - The software component that virtualizes a network function. ● EPC becomes vEPC ● NFVI (NFV Infrastructure) – An Infrastructure that allows us to effectively play around with the VNFs.
What can we Virtualize ? ● IP Functions: ● Network Security and Firewalls ● NAT ● DNS ● Caching ● Intrusion Detection ● Telco Functions : ● Service and Packet Gateways ● Customer Premise Equipments (vRGW, vSTB) ● Evolved Packet Core (vEPC) ● IP Multimedia Service (IMS) ● Mobility Management Entity (MME) ● C-RAN – EnodeB Functions
High level NFV Framework
OPNFV and Software Defined Everything
OPNFV Architecture
SUSE - Software Defined Infrastructure Application Delivery Custom Micro Service Applications Platform as a Service Kubernetes / Magnum Cloud Foundry
Some of the important OPNFV Projects ● FuncTest – Functionality Tests ● Yardstick – Infrastructure compliance ● DPACC – Data Plane Acceleration ● SFC – Service Function Chaining ● Doctor – Fault Management ● IPV6 ● Releng – Release Engineering ● Pharos and XCI – Test Lab and CI
Realizing NFVI Requirements
NFVI Requirements ● Virtualized resources – KVM, Containers ● A tool to manage virtualized resources – OpenStack, Kubernetes ● Scalability ● Availability and Security ● Simplified Service Delivery ● Reliable High Performance
Virtualization - OpenStack ● Designed to scale horizontally ● Fully open source ● Integrates with third-party software and hardware components ● Well-defined APIs to integrate with other NFVI Components ● Several NFVI features integrated in core projects (e.g: Nova, Neutron, Keystone) ● In production environments ● Tested explicitly in OPNFV regularly for functionality and performance
SUSE OpenStack and SLES Virtualized resources KVM Virtual Machines A tool to manage virtualized OpenStack resources Scalability Already in production Availability and Security Live Patching Simple Service Delivery Through AAS and Containers Reliable High Performance Real Time services and tuning
Kubernetes – Why we need it Container orchestration at scale Why Kubernetes ● Containers ● Microservice ● Cloud Native
Kubernetes – What it provides Container orchestration at scale ● Service discovery and load balancing ● Automatic binpacking ● Self healing ● Automated rollouts and rollbacks ● Horizontal scaling
Kubernetes – Check NFVI requirements Container orchestration at scale Virtualized resources – KVM, Yes Containers A tool to manage virtualized Yes resources Reliable High Performance Yes Scalability Yes Availability and Security Yes and limited Simplified Service Delivery Yes
Performance
Reliable, Secure and Consistent Performance There are really just two things that needs to perform well reliably to achieve this: ● Computing ● Networking
Computing Features If you are running everything on common hardware, it all depends on the Operating System ● Real Time Computing – your packet never waits for its time ● Huge Memory Pages – all the data you need in your neighborhood ● vCPU Pinning – do not share your processing power with others ● NUMA Affinity – processing where the data lies ● Live Patching – enhance your security without reboot
Networking Features If you are running everything on common hardware, it all depends on the Operating System ● Software Switch ● PCI Passthrough ● SRIOV ● OVS-DPDK
Traditional Virtualized Networking VM2 VM1 VMM L2 Software Switch NIC
Device Passthrough VM2 VM1 VMM L2 Software Switch NIC1 NIC1
SRIOV VM1 VM2 VMM L2 Software Switch VF1 VF2 NIC
DPDK VM1 VM2 VMM vhost user Open vSWitch + DPDK DPDK PMD NIC
Demo
Demo Content Moving forward towards NFVI ● A sample VNF application with SOC ● Updatability through live security update ● SLES real time kernel scheduling ● Performance tuning with CPU Pinning and NUMA Affinity ● Performance under stress tests
Links for more information https://www.etsi.org/technologies/nfv ● https://wiki.opnfv.org/ ● https://cilium.io ● https://www.dpdk.org/ ● https://fd.io/ ● https://www.onap.org/ ● https://osm.etsi.org/ ● https://www.suse.com/documentation ●
Sources for Images ● Fujitsu EPC, EnodeB ● https://www.fujitsu.com/global/documents/about/resources/publications/fstj/archives ● Cisco CPE, Serving Gateways ● https://www.cisco.com/c/en/us/products/wireless/sgw-serving-gateway/index.html ● 5G Applications ● https://www.etsi.org/technologies-clusters/technologies/5g
Questions ?
Recommend
More recommend