Can Virtual CPE Be Cost-Effective For Enterprise Customers?

Transcription

Can Virtual CPE Be CostEffective for EnterpriseCustomers?Charlie Ashton, Wind River1 2016 Wind River. All Rights Reserved.

Topics Introduction to Wind River in telecom Business drivers for virtual CPE Four key challenges for cost-effective vCPE Service reliability Service agility Virtualization overhead Server footprint Summary2 2016 Wind River. All Rights Reserved.

Introduction to Wind River in NFV Founded in 1981 as embedded tools andOperating System company Acquired by Intel in 2009 Our software has been deployed in over twobillion devices Where failure is not an option Telecom customers include all the top 20 TEMsworldwide3 2016 Wind River. All Rights Reserved. Solving critical challenges for network virtualization Titanium Server: only commercial NFV cloud thatdelivers the Carrier Grade reliability required fortelecom networks Supported by Titanium Cloud ecosystem of industryleading partners

An Open Solution for NFV InfrastructureOSS / BSSNFV OrchestratorVirtual NetworkFunctions (VNFs)VMVMVMVNFVNFVNFControl NodeCompute NodeStorage NodeCarrier Grade CloudManagement and TelcoMiddlewareCarrier Grade Accelerated vSwitchCarrier GradeStorage ClusterOpenStackCarrier Grade KVMCarrier Grade LinuxTitanium Server Software4 2016 Wind River. All Rights Reserved. 100% compatible withopen standards: novendor lock-in Full Carrier Gradereliability for virtualizedservices Maximum VM densitydrives OPEX savings

Titanium Cloud Ecosystem PartnersOSS / BSSNFV OrchestratorsVNFsServer Platforms5 2016 Wind River. All Rights Reserved.

Addressing Multiple Virtual CPE Deployment OptionsSupporting three deployment models planned by service providersTitanium ServerOnly two nodes99.9999% uptimeService Provider Data CenterVNF accelerationService agilityTitanium ServerVNF partnerships supporting allthree deployment models6 2016 Wind River. All Rights Reserved.Service ProviderPoint-of-Presence (PoP)orCentral Office (CO)Titanium Server CPECustomer Premise

Why Virtual CPE for Enterprises?Revenue growth plus cost savings for service providers7 2016 Wind River. All Rights Reserved.

Revenue Growth from Agile, On-Demand Services Managed network services for enterprises isalready a lucrative market for service providers 135B worldwide in 2016 MPLS, Ethernet, VPN, WAN Optimization, Security vCPE presents significant growth opportunity forfirst movers Scalable, automated deployment of existing and newservices Self-provisioning by customers accelerates adoption cycle 5% annual revenue growth per site8 2016 Wind River. All Rights Reserved.

New Revenue from Upsell and Cross-Sell of Services Managed network services enabled byvirtualization Dynamic VPN configuration Expectations set by portals from cloud serviceproviders (Amazon, Google etc.) Dynamic WAN configuration Automated, scalable platforms Self-service bandwidth-on-demand User-friendly dashboards Data center interconnect services On-demand instantiation of CPE VNFs Security as a service (firewall, DDoS etc.) WAN optimization Web acceleration9 On-demand provisioning is key 2016 Wind River. All Rights Reserved.

40-50% Cost Savings from Migration to vCPE Replace expensive, dedicated physicalappliances with standard servers 42% savings in equipment CAPEX and OPEX Remove inefficiencies in manual equipmentinstallation and service provisioning 72% savings in “Order to Cash” (O2C) Eliminate most truck rolls and accelerateremote resolution of failures 61% reduction in “Trouble to Resolve” (T2R) costs10 2016 Wind River. All Rights Reserved.

SDN and NFV are Key Enablers for Cost Savings Equipment CAPEX and OPEX savings driven bylow-cost standard servers Efficient resource utilization thanks to virtualization No stranded assets Aggressive VNF pricing driven by competition SDN and NFV drive efficiency in Order-to-Cash Accelerated, self-service provisioning of new services CPE virtualization minimizes support costs Customer site repairs mostly eliminated11 2016 Wind River. All Rights Reserved.

So . vCPE is Top NFV Use Case for 2015-201612 2016 Wind River. All Rights Reserved.

Four Challenges to Cost-Effective Business CPE Service reliability Service agility Virtualization overhead Server footprint13 2016 Wind River. All Rights Reserved.

vBCPE Challenge #1: Service Reliability Enterprise customers expect high reliability frommanaged services Whether provided by traditional appliances or delivered byvirtual CPE software Typical requirement is five-nines uptime (99.999%) Service downtime impacts service provider revenues Service Level Agreement (SLA) penalties plus operational expenses plus customer churnVirtualized CPE services need to maintain reliability of traditional physical implementations14 2016 Wind River. All Rights Reserved.

IT Cloud Platforms Don’t Deliver Telco ReliabilityIT Platform CapabilityEnterprise vCPERequirementsDetection of failed VM 1 minute 1sDetection of failed compute node 1 minute 1sRecovery from control node failureNo support 25s1-2 GbpsLine rate with minimumcore utilizationDepends on Linuxdistribution50msNo supportFull supportvSwitch performanceNetwork link failure detectionLive migration for DPDK-based VMs15 2016 Wind River. All Rights Reserved.

Service Reliability Requirements Are ComplexVM lifecycle management, software updates, security and manceswitchingLive VM migrationAutomatic failuredetection andrecoveryEnterprise vCPErequirementsAdvanced softwaremanagementVM scale-up andscale-downTelco-grade AAAsecurity16 2016 Wind River. All Rights Reserved.Carrier Gradestorage

One Solution that Meets All the RequirementsIT PlatformsEnterprise vCPERequirementsDetection of failed VM 1 minute 1s500msDetection of failed compute node 1 minute 1s1sRecovery from control node failureNo support 25s 25s1-2 GbpsLine rate with minimumcore utilization40 Gbps with 2 cores(256B packets)Depends on Linuxdistribution50ms50msNo supportFull supportFull support: 200msvSwitch performanceNetwork link failure detectionLive migration for DPDK-based VMs17 2016 Wind River. All Rights Reserved.TitaniumServer

vBCPE Challenge #2: Service AgilityExample OpenStack has no primitives to reconnectthe firewall interface from the router to theWAN acceleratorBefore:RouterFirewall Only options are: Either delete the firewall interface and reconnect,which may lead to ambiguity because firewall rulestied to specific virtual NICAfter:RouterWANAcceleratorFirewall Or provision new service chain from scratch whichcauses outage of at least five minutesNeed a solution for reconfiguring service chains with minimal service downtime18 2016 Wind River. All Rights Reserved.

Two Solutions for Service Chain ReconfigurationOption 1: Orchestrate service chain update usingOpenStack1ServiceA2HEAT StackA Accelerated by use of HEAT stack foreach service1Option 2:ServiceA1ServiceA3HEAT StackB2HEAT StackA Reconfigure vSwitch flows using SDNServiceBServiceBServiceBLANServices(VNFs)HEAT StackC3HEAT StackB2ServiceCServiceC4HEAT StackC3ServiceNServiceNLANHEAT StackNNServiceCLANAdd a new service in seconds vs. weeks or months today19 2016 Wind River. All Rights Reserved.HEAT StackAHEAT StackBHEAT StackNHEAT StackCInitial servicechainAdd new serviceat the end: Change Stack C Create Stack NAdd new servicein the middle: Change Stack C Create Stack N

vBCPE Challenge #3: Virtualization OverheadVNFs Multiple elements contribute to bottom-linecost-per-user Needs to be favorable vs. physical appliancesFixed-FunctionAppliances per userVirtualization Platform Goal is lower costs plus increased revenuesStandard Servers Virtualization overhead is a major factor per userCAPEX, OPEX, Order-to-Cash,Trouble-to-Resolve etc. Off-the-shelf virtual switches have very lowperformance vs. physical switches Limits VNF performance Causes inefficient resource utilizationNeed to ensure that virtualization overhead is minimized20 2016 Wind River. All Rights Reserved.

High Performance vSwitching Drives OPEX Savings Example: Accelerated vSwitch integrated inTitanium Server 40x performance of Open vSwitch (OVS) Fully compatible with Open vSwitch: standard APIs Increased switching performance greater VMdensity Fewer cores required to run vSwitch More cores available for VMsGreater VM density reduced OPEX More users per serverVirtualization overhead is minimized21 2016 Wind River. All Rights Reserved.Example use case: virtualized media gatewaySystem configuration: 28-core platform (dual socket) with one VM per core Bandwidth required: 3.5 Gbps per core (6.8 Mpps per core)Most efficient implementation using Open vSwitch 23 cores required for switching, 1 core running VM, 4 unusedMost efficient implementation using Titanium Server 10 cores required for switching, 17 cores running VMs, 1 unused17x improvement in VM density

vBCPE Challenge #4: Server FootprintEspecially critical issue for customer premise deployments Need to deploy vBCPE on low-cost servers Customers won’t pay more for servers than for physical appliances Solution price has to include VNFs and virtualization platformService Provider Data Center Must ensure service reliability required by enterprises Minimum two servers for redundancy on hardware failures Server utilization must be optimized Only compute nodes run the services that generate revenueService ProviderPoint-of-Presence (PoP) Control and storage nodes represent overhead costsorCentral Office (CO)22 2016 Wind River. All Rights Reserved.Customer Premise

One Small-Footprint Solution: Titanium Server CPEComplete, high-reliability vBCPE on just two servers Compute, control and storage nodes instantiated on each server Only one processor core required for control and one for storage Maximizes cores available for accelerated vSwitch and VMs: revenue Ensures service uptime required by enterprises Six-nines infrastructure reliability enables five-nines services Accelerated vSwitching maximizes number of users per serverTitanium Server CPE Validated, pre-integrated VNFs for complete vCPE solution23 2016 Wind River. All Rights Reserved.Customer Premise

SummaryKey challenges to cost-effective enterprise virtual CPE are all solvable Service reliability Service agility Virtualization overhead Server footprint24 2016 Wind River. All Rights Reserved.Business upside opportunity for serviceproviders is significant and achievable

Please Come and See Our Complete vCPE DemoOrchestrationRouterWind River pod inthe Intel booth25 2016 Wind River. All Rights Reserved.FirewallVirtual CPEWAN Optimization

Addressing Multiple Virtual CPE Deployment Options Supporting three deployment models planned by service providers Customer Premise Service Provider Data Center Service Provider Point-of-Presence (PoP) or Central Office (CO) Titanium Server Titanium Server Titanium Server CPE Only two nodes 99.9999% uptime VNF acceleration Service agility