OpenCloud: Value-Add Cloud Bringing Openness and Innovation to Network and Cloud Infrastructure

What Is Your Cloud Strategy? Directly Take on the OTT Cloud Providers Great long-term goal, but they have a 10-year head start Risk that you are always playing catch-up Apply Cloud Technology to the Central Office Great first step, but too modest as a long-term goal Risk that narrow use case doesn’t pay long-term dividends Answer  Build a Value-Added Carrier Cloud Treat OTT providers as commodity Gain near-term CAPEX/OPEX savings

Take on OTT Providers It’s about services, not just infrastructure Amazon is able to leverage a rich service ecosystem Slide says Amazon, but similar experiences elsewhere (Google, Microsoft, Akamai). Main point – our approach is based on experience.

Central Office as a Data Center NFV Orchestration NFV chaining L2VPN/L3VPN/ L3 Connectivity L2 Connectivity Still managing (virtual) servers; may improve CAPEX but not OPEX Misses opportunity to innovate around network-wide services BNG vSW Cache vSW First have punch-line on complexity – Managing Servers rather than services IDS WAN-Accel EtherSW+ ROADM Commodity Servers & Storage

What Should Be Your Cloud Strategy? Treat OTT Cloud Providers as Commodity Treat Central Office as an Extension of the Cloud Leverage Strategic Infrastructure to Build a Value-Added Cloud that providers better… Responsiveness Bandwidth Privacy Trust Predictability Customization

OpenCloud – Pilot Value-Added Cloud Organizing Principle Everything-as-a-Service (XaaS) Service Composition & Custom PaaS Open Source Software XaaS Operating System (XOS) Network Hypervisor (OpenVirteX) Global Install Cloud (OpenStack-as-a-Service) Operational Deployment 5 “Data Centers” 12 Internet2 Routing Centers 100+ Edge Sites (Regional and Campus Networks) Includes Access to EC2 and Commodity Services

Scalable Function = Service Everything-is-a-Service (XaaS) Unifying Architecture SDN Cloud NFV about best practices in scaling functionality about treating the network’s data plane as a programmable function network’s control plane Scalable Function = Service Everything-is-a-Service (XaaS)

Anatomy of a Service Service Controller Logically centralized interface to network-wide functionality. Service Controller Network-as-a-Service Compute-as-a-Service Content Delivery Network Many points of implementation distributed across the network. Service Instances (VMs, Devices) OpenFlow Switches Virtual Machines Caches

Service Composition Service3 = Service1  Service2 Service1 Service2 Logically compose services to achieve desired functional/security behavior Service1 Service2 VM VM VM VM VM VM VM VM VM VM VM Map onto physical infrastructure to meet performance requirements VM VM VM VM VM VM VM VM VM VM VM VM VM VM

(Mix of own services and 3rd party services) Custom PaaS Platform for Data Analytics Platform for NFV Operations Platform for Service Development … Library calls Library calls … Library calls … Library = Mashup, Script… Factored so all issues of scalability are captured in services; just the “Service Controller S1 S2 S3 S4 S5 Sn Pool of Services (Mix of own services and 3rd party services)

Operational Example Syndicate = CDN  Object Store  NoSQL DB Value-Add Storage Service Scalable Read Bandwidth (Akamai HyperCache & RequestRouter) Data Durability (S3, Glacier, DropBox, Box, Swift) Data Consistency (Google App Engine)

Value-Added Storage Service Amazon AWS Google Cloud Platform … Commodity Clouds Operator Data Centers Backbone PoPs Central Offices Subscribers (Enterprises) Nintendo Wii HPC RR S3 HPC Latency matters Shared state matters Sufficient resources matters MS

XOS – XaaS Operating System Defines Service as a first class object Service = {Slice,…} + Controller-Plugin Flexible control over operational parameters Isolation: Slice = {VM,…} + {VN,…} VM = Placement + Image + Resources (e.g., cores) VN = Topology + NetOS + Resources (e.g., bandwidth) Composition: VN1 | VN2 | VN3 Programming environment for orchestration Specify Service Composition Policy Create Custom PaaS Monitor & Manage Service Scaling S

XOS Software Components OpenCloud Management System XOS SaaS PaaS MaaS CaaS (Nova) NaaS (Neutron) IDaaS (Keystone) MaaS MaaS MaaS MaaS MaaS OpenVirteX XaaS is the organizing principle. All about “Service Orchestration” (managing services, not servers) and not just IaaS Node Node OpenStack OpenStack

Controller Plugin Framework XOS Details Tenant View Developer View Operator View . . . PAL Data Model Controller Plugin Framework Adheres to MVC Paradigm Service 1 Controller Service 2 Controller Service 3 Controller Service 4 Controller Service N Controller . . .

OpenVirteX Network OS Network OS Network OS VM OpenVirteX – A Network Hypervisor that supports Topology, Address Space, and Control Isolation Physical Network Key capability -- Dynamically create customizable Virtual Networks.

Specify Service Composition Virtual Net (e.g., Big Switch) S Service “S” deployed on a scalable set of VMs RR Internet OS HPC Clients MS

Create Custom PaaS … ...

Monitor & Manage Service Scaling

– Operational Deployment on Internet2 – OpenCloud – Operational Deployment on Internet2 –

– Central Office Reimagined as a Data Center – ONOS Use Case 2 – Central Office Reimagined as a Data Center – Amazon AWS Google Cloud Platform … Commodity Clouds Operator Data Centers Backbone PoPs Central Offices Subscribers (Enterprises) Nintendo Wii Latency matters Shared state matters Sufficient resources matters

NF-as-a-Service XOS + OVX + ONOS PGW XCODE NLA CDN Mobile Customers PGW XCODE NLA CDN Residential Customers BNG CDN CG-NAT Firewall Enterprise Customers VPN WanEx DSA IDS XOS + OVX + ONOS OLT Packet SW + ROADM …

NF-as-a-Service Internet S SDN-IP running as an ONOS application Access Subscriber BNG … Subscriber RR AUTH HPC Virtual Net (e.g., Big Switch) OS Wide-Area Acquisition Net running on ONOS S Service “S” deployed on a scalable set of VMs

Soliciting… Early Adopters (Brave Users) Demonstration Services Edge Sites (OpenStack Clusters)