Network Function Virtualization (NFV)

Concepts
 Network Functions Virtualization (NFV):
 Virtualizes network functions using software on VMs
 Replaces proprietary hardware (e.g., NAT, firewall, DNS)
 NFV Benefits
 Decouples hardware & software
 Scalable, flexible resource usage
 Runs on COTS x86 servers
 Virtualized Devices
 Network Functions: Routers, switches, CPE, DPI
 Compute Devices: Firewalls, IDS, management tools
 Storage: Network-attached file/database servers
 Traditional vs NFV
 Traditional: Fixed, closed, hardware-dependent
 NFV: Dynamic, shared, software-driven
NFV Principles
 Service Chaining

 VNFs = Modular blocks

 Traffic flows through multiple VNFs

 Enables desired end-to-end functionality

 Management and Orchestration

 Manages VNF lifecycle: create, monitor, relocate, shut down

 Handles service chaining, infrastructure, and billing

 Distributed Architecture

 VNFs consist of VNFCs (components)

 VNFCs can run across multiple hosts

 Enables scalability and redundancy

High-Level NFV Framework
Three Domains of NFV
 Virtualized Network Functions
(VNFs)
 Software-based NFs running on
NFV
 NFV Infrastructure
 Virtualizes compute, storage, and
network resources
 Management and Orchestration
 Manages VNF lifecycle and NFVI
resources
 Handles virtualization-specific tasks
 VNF Relationships
 VNF Forwarding Graph (VNF
FG):Specifies flow between VNFs
(e.g., firewall → NAT → load
balancer)
 VNF Set: VNFs grouped without
defined connectivity (e.g., server
pool)
NFV Benefits
 Reduced Capital Expenditure (CapEx):
 Commodity hardware, equipment consolidation, pay-as-you-
grow models
 Eliminates overprovisioning
 Reduced Operational Expenditure (OpEx):
 Lower power, space, and management costs
 Simplified network control
 Faster Service Deployment:
 Quick rollout of services
 Reduced risk, faster ROI, agile service evolution
 Interoperability:
 Standardized, open interfaces
 Resource Efficiency:
 Shared platform for multiple users/apps/tenants
 Agility & Flexibility:
 Dynamic scaling to meet demand
 Geo- or customer-specific service targeting
 Open Ecosystem & Innovation:
 Encourages small players and academia
 Enables new services and revenue streams at lower risk
NFV Requirements
 Portability & Interoperability:
 Run VNFs from various vendors on standard hardware
 Decouple software from hardware via standardized interfaces
 Performance Trade-offs:
 Manage potential performance drops due to lack of specialized
hardware
 Minimize latency and overhead with efficient software/hypervisors
 Migration & Coexistence:
 Support hybrid environments (physical + virtual appliances)
 Maintain compatibility with existing interfaces
 Management & Orchestration:
 Unified architecture aligned with standards
 Simplified control and monitoring of VNFs
 Automation:
 Essential for scalability and operational efficiency
 Security & Resilience:
 Maintain network integrity, availability, and resistance to attacks
 Network Stability:
 Ensure stability during VNF relocation, failure recovery, or attacks
 Simplicity:
 Aim for simpler operations compared to legacy systems
 Integration:
 Enable multi-vendor compatibility without high costs
 Ecosystem must support validation and third-party maintenance
NFV Reference Architecture
NFV Reference Architecture
 4 Key Components
 NFV Infrastructure (NFVI):
 Virtualizes compute, storage, and network resources into
resource pools.
 VNF/EMS:
 Software-based network functions (VNFs) + Element
Management Systems (EMS).
 NFV-Management and Orchestration:
 Manages and orchestrates VNFs and infrastructure
resources (compute, storage, network).
 OSS/BSS:
 Business and operational support systems.
 Three Architectural Layers
 Infrastructure Layer:
 NFVI + Virtualized Infrastructure Manager (VIM)
 VNF Layer:
 VNFs, EMS, and VNF Managers
 Management & Orchestration Layer:
 OSS/BSS + NFV Orchestrator
NFV Reference Architecture
 NFV-Management and Orchestration Functional Blocks:
 NFV Orchestrator:
 Manages NS/VNF lifecycles, global resources, and
authorization.
 VNF Manager:
 Handles VNF instance lifecycle.
 Virtualized Infrastructure Manager (VIM):
 Manages virtual/physical resource interaction.
 Key Reference Points (Interfaces):
 Vi-Ha: Interface to physical hardware
 Vn-Nf: VNF to virtual infrastructure API
 Nf-Vi: NFVI to VIM
 Or-Vnfm: Orchestrator to VNF Manager
 Vi-Vnfm: VNF Manager to VIM
 Or-Vi: Orchestrator to VIM
 Os-Ma: Orchestrator to OSS/BSS
 Ve-Vnfm: VNF lifecycle management
 Se-Ma: Access to deployment templates & infrastructure models
NFV Reference Architecture
NFV Implementation Overview
NFV Reference Architecture
NFV Implementation Overview
 Standards & Open Source:
 ISG NFV:
 Developing standards for NFV interfaces and components.
 OPNFV (Open Platform for NFV):
 Launched by Linux Foundation (2014) to accelerate NFV
adoption.
 Key Objectives of OPNFV:
 Build integrated, tested open source NFV platform
 Ensure operator-driven validation of releases
 Contribute to & align with relevant open source projects
 Foster open NFV ecosystem (standards + software)
 Promote OPNFV as preferred open reference platform
 Initial Focus of OPNFV
 Development of:
 NFV Infrastructure (NFVI)
 Virtual Infrastructure Manager (VIM)
 APIs to Management and Orchestration and VNFs
 Provides common base for vendors to build VNF &
Management and Orchestration solutions
NFV Infrastructure

Core Domains
 Compute Domain
 Commercial-Off-The-Shelf (COTS) servers and storage (high-
volume, standard hardware)
 Hypervisor Domain
 Abstracts hardware for VMs
 Mediates compute resources for VNFs
 Infrastructure Network Domain (IND)
 High-volume switches forming configurable network
 Delivers infrastructure-level network services
Container Interface in NFV (European Telecommunications
Standards Institute (ETSI) Perspective)
Container Interface in NFV (European Telecommunications
Standards Institute (ETSI) Perspective)

 Important Clarification
 ETSI states that "Container Interface" ≠ Container
Virtualization .
 Interface Types
 Functional Block Interface:
 Connects two software blocks (can be on different hosts)
 Enables inter-block communication
 Container Interface:
 Execution environment on a single physical host
 Hosts and runs a functional block locally
 Key Features:
 Highlights execution on physical hosts
 Differentiates communication vs. execution environment
 Crucial for understanding VM/VNF behavior within NFVI
Container Interface in NFV (European Telecommunications
Standards Institute (ETSI) Perspective)
 ETSI NFVI Architecture: Key Insights
 VNF architecture is separate from the hosting NFVI architecture.
 VNFs and NFVI consist of distinct domains (compute, hypervisor,
network).
 Management & Orchestration (MANO) is a separate domain, but
often overlaps with NFVI (e.g., element management).
 Container Interface:
 Connects VNFs to NFVI
 Hosts VNF & MANO functions as VMs on the same physical host
 Three-Layer VNF Deployment
 Physical Resources
 Virtualization Layer
 Application Layer (VNFs)
→ All typically co-located on the same physical host.
 Interface Types (As shown in Figure)
 Container Interfaces: Same host (Interfaces: 4, 6, 7, 12)
 Functional Block Interfaces: Cross-host or distributed (Interfaces:
3, 8, 9, 10, 11, 14)
 Legacy Network Interfaces: For integration with non-NFV networks
(Interfaces: 1, 2, 5, 13)
 NFV vs. SDN
 NFV: VNFs run on same host as virtualization software
 SDN: Control & data planes are often on separate physical hosts
Deployment of NFVI Containers

 VNF Components (VNFCs) and Deployment

 Single Host Setup (Fig. a)
 One VM per VNFC
 Multiple VMs run on one host via a hypervisor
 Each VM = 1 VNFC
 Hosted on compute container interface
Deployment of NFVI Containers

 VNF Components (VNFCs) and Deployment

 Distributed Setup (Fig. b)
 Multiple VNFCs form one VNF
 VNFCs can run across different compute nodes
 Nodes interconnected via infrastructure network domain
Logical Structure of NFVI Domains

NFVI Logical Structure (ISG NFV)

 ISG NFV standards define the logical architecture of NFVI domains
and interconnections.
 Supports both open source and proprietary implementations.
 Provides a framework for development and identifies key interfaces
between components.
Compute Domain
 CPU/Memory: COTS processor and main memory for executing
VNFC code.
 Internal Storage: Local nonvolatile storage (e.g., flash).
 Accelerators: Optional hardware for security, networking,
packet processing.
 External Storage: Access via storage controller to secondary
memory.
 NIC: Connects compute node to infrastructure network (Interface
14 – Ha/CSr-Ha/Nr).
 Control/Admin Agent: Interfaces with VIM (see Fig. 7.8).
 Eswitch: Server-embedded switch, functionally part of
infrastructure network.
 Execution Environment: Presented to hypervisor by
server/storage (Interface 12 – VI-Ha/CSr).
Eswitch

 Two VNF Workloads:

 Control Plane: Protocols like BGP, CPU-intensive, light I/O.
 Data Plane: Routing/switching tasks & high I/O demands.
 Challenge in Virtualized Environment:
 Traffic goes through hypervisor’s virtual switch.
 Adds processing overhead and latency.
 Eswitch Solution:
 Bypasses hypervisor.
 Enables direct memory access (DMA) to NIC.
 Boosts performance with zero processor overhead.
NFVI Implementation Using Compute
Domain Nodes
 VNF Structure & NFVI Nodes
 VNFs: Made of one or more VNFCs (VNF Components)
 VNFCs: Run as software on VMs via hypervisors on compute
nodes
 Virtual Links: Defined via Infrastructure Network Domain
(IND)
 NFVI Node: Group of physical devices managed as one,
supports VNF execution
 Types of Compute Domain Nodes
 Compute Node: Executes fast, deterministic instructions
 Gateway Node: Connects NFVI to transport & legacy
networks
 Storage Node: Offers storage via local or remote access (e.g.,
NFS, Fibre Channel)
 Network Node: Provides switching/routing using compute &
storage
 NFVI-PoP (Point of Presence)
 Multiple physical devices per compute domain
 Distributed locations enable service diversity
NFVI Implementation Using Compute
Domain Nodes
 Deployment Scenarios
 Monolithic Operator – Single org. owns hardware and VNFs
(e.g., private cloud)
 Operator Hosting VNOs – Host multiple operators (e.g.,
hybrid cloud)
 Hosted Network Operator – IT org runs infra; operator runs
VNFs (e.g., BT, Verizon)
 Hosted Communication Providers – Multiple providers
hosted (e.g., community cloud)
 Hosted Communication + Application Providers – Add
public app hosting (e.g., public cloud)
 Managed Service on Customer Premises – Provider
equipment at client site
 Managed Service on Customer Equipment – VNFs on client-
owned hardware
Hypervisor Domain

 Abstraction Layer: Manages hardware for VM operations (start,

stop, scale, migrate)
 Main Components:
 Compute/Storage Management : Virtualized access to VMs
 Network Management: Virtualizes NICs for VMs
 VM Management & API: Supports VNFC execution (Interface
7)
 Control/Admin Agent: Interfaces with VIM
 vSwitch: Virtual Ethernet switch connecting virtual Network
Interface Cards of VMs
 vSwitch Operation
 Same Host: VNFs connect via local vSwitch
 Different Hosts: Traffic flows → vSwitch → NIC → external
switch → target NIC → vSwitch → VNF