NFV Guide - Ixia

DEMYSTIFYING
NFV
IN CARRIER NETWORKS
A Denitive Guide to Successful Migrations
Pierre Lynch, Michael Haugh, Liza Kurtz, & Joseph Zeto. (1st ed.)
Demystifying NFV
in Carrier Networks:
A Definitive Guide to
Successful Migrations
Copyright 2014 Ixia. All rights reserved.

This publication may not be copied, in whole or in part, without Ixias consent. Ixia, the Ixia logo, and all
Ixia brand names and product names in this document are either trademarks or registered trademarks
of Ixia in the United States and/or other countries. All other trademarks belong to their respective
owners. The information herein is furnished for informational use only, is subject to change by Ixia
without notice, and should not be construed as a commitment by Ixia. Ixia assumes no responsibility or
liability for any errors or inaccuracies contained in this publication.
Demystifying NFV in Carrier Networks
Table of Contents
Preface 1
Why Read This Book?
About Ixia
Introduction: Everything Known is Unknown Again

Validating Real Performance in a Virtual World
Chapter 1: What is NFV?

1.1
6
10
A Move Toward Software, Simplicity, and Standardization
10
1.2 The Market for NFV
17
1.3 Vendor Milestones
18
1.4 Carrier Milestones
20
1.5 NFV in the Mobile Core
21
1.6 Why Brave Another Paradigm Shift?
22
Chapter 2: Drivers for NFV
24
2.1 Increased Service Agility and Flexibility
25
2.2 Economic Advantages
27
2.3 Customers Benefit as Well
28
Chapter 3: NFV Risks and Challenges
30
3.1 Strategic Challenges
31
3.2 Architectural/Implementation Challenges
35
Chapter 4: Validation Strategies

4.1
Virtual Testing vs. Testing Virtualization
43
44
4.2 Critical Components of Validation
47
4.3 Key Performance Indicators (KPIs)
48
4.4 Scope of Validation
49
Table of Contents
Chapter 5: Validating the NFV Infrastructure
53
5.1 Hardware
53
5.2 The vSwitch
54
5.3 The Hypervisor
55
5.4 VM Manager
56
5.5 Management and Orchestration
57
5.6 Virtual Machines (VMs)
58
5.7 Real-world Scenarios
58
Chapter 6: NFV Test Cases
60
6.1 Basic Test Setup
60
6.2 Validating the Virtual EPC
64
6.3 Testing a Network with a Combined Legacy and Virtualized EPC 66

6.4 Testing Virtualized Elements Within the EPC
68
6.5 Testing IMS Virtualization
69
6.6 Testing Virtualized Broadband Remote Access Server

(BRAS) Functionality
71
6.7 Beyond Migration
71
Chapter 7: Maintaining Visibility: Monitoring, Access,

and Control in a Virtualized Environment
73
7.1 Virtual Visibility: Evolving Goals

and Best Practices
76
7.2 Phantom Taps and the New Virtual

Visibility Framework
77
Chapter 8: Ixia NFV Test Solutions
80
8.1 Comprehensive Real and Virtual Test Capabilities
80
8.2 The Most Trusted Names in Networking Trust Ixia
83
Acronyms and Terms
84
Table of Contents
Demystifying NFV In Carrier Networks
Preface
Network Functions Virtualization (NFV) is one of the most profound
paradigm shifts the networking industry has faced to date. Proven
functions such as routing, policy, firewall, DPI, and many others will
move from running on dedicated hardware appliances to running
on unproven virtualized server platforms in the hope of achieving
massive efficiencies.
While carriers agree on the need and vision for NFV, many are
struggling to quantify the benefits, understand the practical
migration steps, and measure success:
How
should they go about replacing their current network

architecture, moving from purpose-built physical devices to a
virtualized, software-controlled scenario?
How
will they know with certainty that the migration was

successful and justifies their investment?
Does
the virtualized solution deliver the performance,

resiliency, security, elasticity, and quality of experience (QoE)
needed to satisfy customers?
With network usage, user expectations, and competitive threats

all on the rise, the risks to operators brands and bottom line are
too great to leave NFV deployments to chance. Everything known
becomes unknown again, and there are risks associated both with
moving too fast, and not fast enough.
From beginning to end, targeted test strategies and methodologies
are needed to accelerate and ensure the delivery of NFV solutions
Preface
with guaranteed quality. Ixia solutions validate that the motivations

for NFV are achievable prior to migration, and that the expected
benefits are realized in the virtualized environment. We help
operators understand when it makes sense to virtualize an aspect
of the network and when it doesnt, delivering clear insight into the
migration process and its ultimate success.
Demystifying NFV
Reduced
Vendor
Lock-In
Elasticity
Agility
Service
without
Impact
Cost
Savings
Promises
Reality
Expertise
Performance
Security
Visibility
Reliability
The promises of NFV are vast and enduring, but operators face many
diverse challenges in ensuring the same or better performance,
reliability, and security achieved by traditional infrastructures.
Preface
Why Read This Book?

Demystifying NFV helps fast-track planning and deployment
by exploring:
What
NFV is and what changes
Virtualization
Benefits
market dynamics and migration forecasts
of adopting NFV
Deployment
challenges
Evolving
best practices for evaluating and validating NFV

strategies and initiatives
High-level
test cases for NFV validation
Well begin with a bit more background on what network

functions virtualization is and the potential benefits it stands to
deliver, then explore the obstacles to deployment and strategies for
overcoming them.
About Ixia
Ixia is an active participant in the European Telecommunications
Standards Institute (ETSI) Industry Specification Groups (ISG) for
NFV, a chair on the Open Networking Foundation, and remains
involved in other leading industry organizations focused on
virtualization. We continue to work with equipment manufacturers
and service providers at the forefront of virtualization to define best
practices for a smooth and profitable migration.
Preface
As paradigm shifts become the norm for mobile operators, our

unrivaled experience working with leading providers worldwide
helps both in transforming networks and fast-tracking the delivery
of compelling new services using virtualized infrastructures. Ixias
unique, comprehensive portfolio of hardware and virtualized test
and monitoring solutions deliver the lab-to-live insights needed
to accelerate and maximize the benefits of NFV throughout the
deployment life-cycle.
Our leadership in virtualization performance testing and optimization
equips service providers to approach and implement Software
Defined Networking (SDN) and NFV with the greatest possible
confidence, efficiency, and support.
Related eBooks From Ixia

For additional insight into major transformations occurring in
service provider networks worldwide, explore other timely
Ixia resources:
Validating VoLTE: A Definitive
Guide to Successful Deployments
Small Cells, Big Challenge: The
Definitive Guide to Designing and
Deploying HetNets
Authoritative Guide to Advanced
LTE Testing
Preface
Introduction
Introduction: Everything Known is

Unknown Again
Driven by providers ceaseless quest to deliver and monetize new
services, mobile network infrastructures are evolving faster than at
any time in history:
Demand
All-IP
continues to grow exponentially
infrastructures have become a reality
Radio
access networks are diversifying as HetNets emerge

introducing small cells and carrier Wi-Fi
5G
is on the radar screen
Amidst this rampant change, the

NFV, done correctly, will
most transformative shift underway
be transformational.
is one designed to speed and
Heavy Reading
facilitate change itself. Network
Functions Virtualization (NFV),
basically the shift of network control to software to gain agility
and a host of operational and economic advantages, is quickly
moving forward.
In tandem with SDN, the migration of networking functions to a
virtual or cloud-based architecture promises to play a pivotal role in
operators ability to innovate new services, accommodate growth,
and compete profitably into the future.
Introduction
Wide-ranging benefits include both technological and

economic advantages:
Accelerating
Increasing
Simplifying
the introduction of compelling new services
network and service agility

network management and optimization
Improving
the economics of service delivery with extensive

capex/opex savings
Proof-of-concepts (PoCs) are underway worldwide and analysts

foresee NFV spending growing at a CAGR of 46% between 2014 and
2019.1 But the risks are great, and the challenges formidable, and still
somewhat unknown.
NFV effectively means having to create and manage a carrier
grade cloud. And where the capabilities and performance of
physical devices are well known and understood, virtualizing various
network functions renders these aspects unknown once again.
Throughout the process, operators must:
Transform
the unknown into the known by validating new

architectural components, assessing the performance of
traditional functions getting moved around, and cultivating new
teams, vendor relationships, and skill-sets
Adopt
an infrastructure model developed for data centers

versus carrier infrastructures and requirements
Invest
in and fast-track PoCs while standards are still evolving
Leverage
evolving best practices to validate decisions, weigh

potential benefits against tradeoffs, and measure success
1 http://www.mindcommerce.com/Publications/NFV_BusCaseMarketAnalysisForecasts_2014-2019.php
Introduction
Validating Real Performance in a Virtual World

With NFV, the old approach to making decisionsrelying heavily
on performance data provided by manufacturersis not only
impractical but impossible because real-world data
does not exist yet. In Chapters 4-7, well offer some
strategies and methodologies for approaching and validating
vital decisions including:
Evaluating
Replicating
new devices and architectural models

the complexities of virtualized wireless networks
Measuring
and guaranteeing subscriber quality of

experience (QoE)
Simulating
security attacks
Monitoring
live virtual networks
Lab-to-live strategies are needed to quantify and evaluate

performance, move forward with confidence, and maintain
quality as traditional and virtualized networks coexist for
the foreseeable future. But lets take a step back and briefly
summarize what NFV is, why operators want it, and the
obstacles they face in making it happen.
Introduction
What is NFV?

In this chapter well take a brief look at what Network Functions
Virtualization (NFV) is, its relationship to Software Defined
Networking (SDN), projected growth, and early efforts by network
equipment manufacturers (NEMs) and operators to date.
1.1 A Move Toward Software, Simplicity,

and Standardization
As is commonly known, the NFV concept was first formally outlined
in October 2012 in a white paper published by an ETSI industry
specifications group. The Network Functions Virtualization ISG that
initially consisted of representatives from thirteen service providers
worldwide has grown steadily in size and scope. More than 150
organizations now participate, and several subsequent papers have
addressed NFV use cases and implementation.
Conceptually, NFV marks a fundamental shift in the service provider
network model in which diverse networking functions move to a
virtual or cloud-based architecture. In practice, NFV may entail:
Migrating
networking functions from proprietary, specialized

hardware appliances to commercial off-the-shelf (COTS) x86based servers
Abstracting
and shifting control of networking functions from

hardware to software by introducing a hypervisor layer
Flexibly
distributing functionality across desired locations

data centers, network nodes, customer premisesto maximize
operational efficiency and performance
Creating
a more application-aware network
10
Targets for NFV include everything from switching, routing, and

broadband remote access servers (BRAS) to access devices to load
balancing and critical security elements such as firewalls and deep
packet inspection (DPI). In mobile networks, NFV represents an
end-to-end proposition reaching from the IP multimedia subsystem
(IMS) and evolved packet core (EPC) to radio access networks
(RANs) to the customer premise.
While virtualization will likely occur on a data center by data center
basis initially, the ultimate vision allows for virtualized network
functions (VNFs) to run in the cloud over logical and geographically
dispersed topologies to achieve maximum efficiency and the lowest
possible total cost of ownership (TCO). The figure below depicts the
evolving NFV architecture.
High-Level NFV Framework
Virtual Network Functions (VNFs)
VNF
VNF
VNF
VNF
VNF
NFV Infrastucture (NFVI)

Virtual Compute
Virtual Storage
Virtual Network
NFV Management
and Orchestration
Virtualization Layer
Compute
Storage
Network
Hardware Resources
11
The proposal is for VNFs to run on COTS hardware such as Dell/

HP servers with standard x86-based computing architectures. The
new virtualization layer shown in the figure above is a hypervisor
that provides virtual access to underlying compute resources,
enabling features like fast start/stop of virtual machines (VMs),
snapshot, and VM migration.
Hypervisor software is able to manage several guest operating
systems and enable consolidation of physical servers onto a
virtual stack on a single server. CPU, RAM, and storage are flexibly
allocated to each VM via software.
Management and orchestration functions will undergo a profound
transformation as this layer must interact with both virtualized server
and network infrastructures, often using OpenStack protocols, and in
many cases SDN. This highest layer is referred to as the Operations
Support System/Business Support System (OSS/BSS), a critical
component of, and perceived obstacle to the widespread adoption
of NFV.
Ultimately, cloud-based management and orchestration stands
to facilitate management of large, highly-distributed network
infrastructures and innovative service offerings. The ETSI ISG
is also proposing an architecture for Service Chaining whereby
multiple network functions are strung together, typically interconnected by a vSwitch.
A virtual network function forwarding graph, or VNF FG can be
created, scaled, and updated very quickly and efficiently. Virtualized
functionality can be added to service chains by instantiating a VM
and simply updating the forwarding graph.
12
As shown in the diagram below, functions can be nested instead of

linear, and stitched together from different physical locations using
various virtualization solutions such as overlay (VXLAN, NVGRE, STT)
and SDN/OpenFlow.
End-to-End Network Service
VNF-FG-2
Endpoint
VNF
VNF
VNF
VNF
VNF
Endpoint
Virtualization Layer
Hardware Resources in Physical Locations

Legend:
NFVI-PoP
Virtualization
Physical Link
Logical Link
Example of an end-to-end network service with VNFs and nested forwarding graphs
With NFV, operators gain greater flexibility in defining service chains

such that a particular service follows the path most applicable to it.
13
1.1.1 NFV and SDN

In honing their virtualization strategies, operators will likely evaluate
and leverage NFV and SDN in tandem. SDN enables the decoupling
of the control plane from the data plane, offering increased
programmability. The ability to program the network through
software in turn promises simplified traffic management and
greater ability to define the way in which packets are forwarded by
networking elements.
The shift away from
distributed protocols
such as Border Gateway
Services
Protocol (BGP) and Open
Shortest Path First (OSPF)
toward more centralized
control (OpenFlow is the
leading SDN protocol
NFV
SDN
thus far with support by
more than twenty leading
vendors) allows various
types of equipment from
multiple vendors to be more easily monitored and managed. As
is a goal with NFV, the more granular control enabled by SDN
allows services to be created and deployed across networks with
reduced equipment requirements. Increased visibility further equips
operators to deliver a higher quality of experience.
14
Industry experts resoundingly agree that NFV and SDN are highly
complementary strategies, though services can be built directly
using NFV without SDN, or be built using SDN without migrating
functions to the cloud. Combining the two stands to deliver
compounded benefits, however, as SDN can be used to provision
network connectivity to VNFs, allowing end-to-end services to be
built with enhanced virtualized functions.
Several top use cases for OpenFlow-enabled SDN and NFV
can be seen in the Open Networking Foundation document
OpenFlow-enabled SDN and Network Functions Virtualization
(February 17.2014).1
Big Picture View of NFV with SDN and Management
Apps
Orchestration
Controller(s)
Controller(s)
Data Center
Wireless
Network
Storage
WAN
Network
Access
Network
Optical
Network
DC Network
Fabric
Virtual
Storage
Server
Virtual Machine
Virtual Machine
Server
Virtual Machine
Virtual Machine
Source: Alcatel-Lucent2
1 https://www.opennetworking.org/images/stories/
downloads/sdn-resources/solution-briefs/sb-sdn-nvf-solution.pdf.
2 Network Functions Virtualizations: Challenges and Solutions, Alcatel-Lucent, 2013.
15
1.1.2 The Evolution of NFV

Late last year, an industry-wide call was made for public PoC
demonstrations with some twenty-five providers banding together
to define progress and processes. Since then, several have
embarked on PoCs.
PoCs are organized by the ETSI ISG for NFV and must include
a service provider and two or more NEMs. Eighteen have been
initiated to date.
Visions for the adoption of NFV vary, as individual operators plans
undoubtedly will. In the diagram below, Virtualized network
functions are those running on hypervisors and general-purpose
hardware but potentially requiring dedicated physical resources
and customized configurations. The Cloud refers to virtualizationbased functions deployed on standard interfaces, while Automated
Life-cycle Management refers to the eventual scenario in which
service providers use tools similar to those found in the IT world to
manage the VNF life-cycle.
NFV Deployment Evolution
Operations Evolution
Automated
Lifecycle
Management
Fully
Automated
Cloud
Auto-Optimization
Cloud
Virtualized
Infrastructure Evolution
Source: ALU
16
As described by ALU, Auto-optimization describes the stage in

which VNFs are able to dynamically and automatically scale to
match available resources with changing demand.
1.2 The Market for NFV

As typically occurs in nascent markets, growth forecasts are all
over the map:
Mind
Commerce estimates global spending on NFV solutions

will grow at a CAGR of 46% between 2014 and 2019. NFV
revenues will reach $1.3 billion by the end of 20193
The
DellOro Group thinks the market could represent $2

billion in equipment sales by 20184
Doyle
Research envisions a best-case scenario in which

NFV gains rapid momentum reaching a $5 billion market by
2018 (including software, servers, and storage)5
As always, its a matter of what aspects of the process each unique

projection encompasses, and how quickly the industry can overcome
the challenges described below in Chapter 3. Of greater and more
immediate importance to operators is how NFV itself will evolve.
This year, Infonetics believes many operators will move
from PoCs to collaborating with vendors to develop and
produce software solutions that will furnish a foundation for
commercial deployments.6
3
4
5
6
www.mindcommerce.com/Publications/NFV_BusCaseMarketAnalysisForecasts_2014-2019.php
www.sdncentral.com/news/nfv-market-size-2b-first-guess/2014/04/
www.lightreading.com/forecasting-the-nfv-opportunity/a/d-id/705403
2014 SDN and NFV Strategies: Global Service Provider Survey. Survey of worldwide service providers
controlling more than 50% of global telecom capex and 47% of revenue
17
1.3 Vendor Milestones

New and traditional vendors are bringing diverse architectures and
products to market:
Alcatel-Lucent
has stated that its working with some 20

service providers on NFV-related initiatives. ALU launched a
portfolio of virtualized mobile network function applications
spanning the EPC, IMS, and radio access network (RAN).
The provider says several network functions are already

available in a virtualized form, with others targeted near-term.
ALU has also stated plans to launch a consultancy practice
aimed at helping operators navigate the transition.
Broadcom
has also been aggressive, working with ARM to

develop a server-class, 64-bit system on a chip (SoC) optimized
for NFV that may ship in volume in 2015. Its Open NFV platform
leverages open-source components such as Linux, KVM
virtualization, and Open Virtual Switch to create a platform that
is not reliant on a single-chip architecture in order to increase
portability and enable the creation of NFV applications that are
highly portable from one platform to another.
Cisco
has been showcasing its Evolved Services Platform,

a unified virtualization and orchestration software platform
designed to equip operators to automate and provision
services across compute, storage, and network functions
in real time. According to Cisco, the platform facilitates
virtualization across a carriers enterprise architecture,
including cloud, video, mobile, and fixed networks.
18
Hewlett-Packard
recently launched an open, standardsbased reference architecture. OpenNFV aims to provide a

complete architectural approach across servers, storage, and
networking to reduce costs. The company has launched a
business unit dedicated to NFV, and introduced OpenNFV Labs
and an OpenNFV Partner Program to speed development of
NFV-based applications.
Huawei
has proclaimed NFV a main focus for the coming

years with offerings including the CloudEdge solution designed
for mobile broadband networks (MBB). CloudEdge purports to
help operators automate network management functions and
speed time-to-market for new machine-to-machine, mobile
video, and mobile enterprise applications.
Intel
has teamed up with Red Hat on an ETSI-approved PoC

for a virtualized EPC.
Juniper
unveiled a suite of SDN and NFV technologies

designed to complement its Contrail SDN controller at Mobile
World Congress 2014. Its NorthStar SDN controller also
targets new network management software for automating
the control of routers and other networking devices, as well as
optical and mobile devices, from a single management plane.
NECs
vEPC solution targets two core network functions:

vMMEs (virtual Mobility Management Entities) and vS/P GWs
(serving and PDN gateways). Virtual network functions are
supposedly decomposed into elementary virtual machines.
Wind
River launched Carrier Grade Communications Server,

an NFV platform intended to help carriers migrate their existing
networks to NFV architectures with minimal disruption and
leave them better able to deploy applications out of the box.
19
1.4 Carrier Milestones

Infonetics has called 2013 the year of the PoC with operators,
manufacturers, and software providers all beginning to wrestle with
deployment issues such as complexity, cost, benefits, and tradeoffs
in their labs. Early efforts have demonstrated some benefits
with regard to scaling traffic and the allocation of resources in
accordance with demand.
This year, the firm expects serious lab trials to progress to
aggressive field trials with commercial deployments debuting in
earnest in 2015 and becoming widespread by 2016. Among the
early leaders are:
AT&Ts
launch of the next generation of its Supplier Domain

Program Domain 2.0 in 2013, targeting migration to
modern, cloud-based architectures. Calling the approach a
transformative initiative, AT&T expects the project to leverage
both NFV and SDN to accelerate time-to-market for advanced
products and services. Initiatives target virtualizing the EPC
to create a multi-service user-defined network cloud
supporting a wide range of network functions and services.
British
Telecom (BT) co-founded and remains active in

the ETSI NFV ISG and claims to be the first network operator
to publish PoC results. BT has engaged in several NFV PoCs
with various partners. Projects to date include:
Testing and successfully deploying a virtualized BRAS
(vBRAS) solution
Testing and deploying virtualization of IPSec tunnel termination

for Wi-Fi and LTE services
Testing virtualizing content distribution networks (CDNs)
20
Colt,
another original member of the NFV ISG, has several

projects underway to leverage NFV and SDN including L3
CPE virtualization and WAN SDN in the data center. External
projects include Colt Live targeting the media sector, elastic
bandwidth, and a Service Console and Dashboard for
enhanced self-service and reporting.
Deutsche
Telecom (DT) has embarked on testing and deploying

its Terastream Architecture in Croatia (based on NFV and
SDN), which includes virtualizing DHCP, IPv4, and VPNs. The
company has also tested distributing BRAS/MBG functionality
performing authentication with SDN, and is discussing
migrating the set-top-box to the cloud.
NTT
DOCOMO has announced plans to commercially deploy

services using a virtualized mobile network during 2016. The
carrier began collaborating with ALU, Cisco, and NEC on NFV in
2013 and recently announced the completion of PoC trials verifying
feasibility of a virtualized EPC.
1.5 NFV in the Mobile Core

With regard to the evolution of NFV in general, a recent white paper
published by Heavy Reading leveraging data from Radisys predicts
that Phase I of the NFV/SDN transformation will be completed
within the next five years.
In terms of the mobile core specifically, Infonetics asked operators
about initial target applications. Fifty-nine (59%) of those responding
for NFV reported plans to deploy mobile core/vEPC by 2016 or
later, and many reportedly plan to leverage a vIMS core for VoLTE
deployment as well.
21
The Evolution of NFV

SDN
NFV
Begin virtualization of control plane

Spread applications freely amongst
geographically-diverse data centers
Level of Network Transformation
Consolidate control planes on a single platform

Virtualize 1-2 applications on the same
core or processor
Separate control plane to dedicated platform
core or processor
Select second app for control/data separation
core or processor
Implement separation of control
& data planes for targeted app in
a single platform
Implement virtualization framework
Identify apps for SDN
Identify control plane
apps for NFV
3
Steps per Year
Source: Radisys7
1.6 Why Brave Another Paradigm Shift?

However fluid and subjective growth forecasts might be, NFV
appears to be real and imminent, though it also looks to be a whole
of work. For all three to be true at once, the prospective benefits
need to be near-term, and far-reaching.
Well take a look at the drivers for virtualization first, then at the
challenges, and evolving best practices for making sure it works.
7 Software-centric Networks: A Migration Path to NFV, Heavy Reading, October 2013
22
Drivers for NFV

NFV and SDN figure to preserve and grow the profitability of
services into the future. The industry is essentially taking a page
from the IT world, where the virtualization of servers
has enabled:
Greater
service agility and programmability including the

ability to program the delivery of services through open APIs
Increased
scalability and elasticity with the cloud-based model

enabling infrastructure as a service (IaaS), platform as a service
(PaaS), and software as a service (SaaS) offerings
Cost-savings
through the consolidation and improved

management of servers while enabling a pay-as-you-go
model
In virtualizing networking functions, operators are seeking

essentially these same benefits. The presumption is that, de-coupled
from the hardware layer, telecommunications network applications
can run on lower-cost, standards-based hardware with far greater
flexibility and efficiency.
Executed correctly, NFV represents greater ability to roll out new,
more compelling and profitable services faster, and with guaranteed
quality. Increased efficiency and savings figure to prove even
more valuable in mobile networks where margins have historically
been tightest and competitive threats from over-the-top (OTT)
providers continue to mount.
24
The virtualization of network functionality stands to deliver two

basic benefits: Increased service agility and substantial costsavings. Each involves many components, representing many longterm advantages.
2.1 Increased Service Agility and Flexibility

Ultimately, service agility translates into the ability to innovate and
launch services faster and more cost-effectively. The evolution from
proprietary equipment to a software-based environment:
Simplifies
provisioning, allowing operators to more quickly

instantiate, move, and evolve services.
Adds
network elasticity, increasing their ability to scale up

and down to meet demand.
Increases
automation, which in turn simplifies operations.

Provisioning for virtual appliances needs to be automated
in order to address the dynamic NFV environment. Doing
so reduces provisioning and configuration times along with
manually induced configuration errors.
Enables
more fluid and efficient resource allocation. Working

in software facilitates specification of CPU, memory, and
other resources, allowing providers to vary allocation and use
according to specific use cases and changing needs. This in
turn enables more fluid capacity management to accommodate
flux in demand, failover, and the like. Higher resource
utilization also results in less equipment being required.
Requires
less specialization as operators can increasingly

leverage tools and skill-sets widely found in IT.
25
Allows
greater proximity to popular IP services.

Virtualizing networking functions allows services to be
provisioned closer to cloud-based services like Facebook
or Pandora, potentially enabling better performance and
economic benefits.
NFV Deployment Drivers
Scale services up or down quickly
86%
Use software for quick revenue
69%
Drivers
Use commercial servers, not network equipment

Operational efficiencies
62%
59%
Multi-tenancy
45%
34%
Real-time network optimization
28% Save energy consolidating workloads

14% VNFs from small players
Percent of NFV Respondents Rating 6 or 7

Source: Infonetics1
Flexibility extends to certain aspects of marketing as well. As the

Infonetics study discusses, providers can use both SDN and NFV to
cost-effectively test new services before committing to full-blown
rollouts. The ability to quickly tweak offerings based on early user
feedback helps in bringing more compelling and profitable services
to market faster, with a lower upfront investment and greater
confidence that theyll succeed.
1 2014 SDN and NFV Strategies: Global Service Provider Survey. Survey of worldwide service providers controlling more than 50% of global telecom capex and 47% of revenue
26
2.2 Economic Advantages

Predictably, another pervasive driver of virtualization is its potential
to deliver higher return on investment (ROI) near-term, and lower
TCO long-term. NFVs economic advantages include significant
reductions in both capital and operating expenditures.
2.2.1 Capex Reduction

The most obvious capital savings stems from trading costly
proprietary hardware for lower-priced commercial platforms. A less
obvious but far greater benefit is not having to invest as heavily in
redundant backup devicesMMEs, SGWs, PGWs, and the likethat
largely go unused because of forced redundant configurations.
Migration to a distributed cloud architecture allows operators to
deploy backup in software using an N+1 configuration, versus having
to invest in and warehouse lots of extra equipment.
2.2.2 Opex Reduction

While the capex savings may feel more immediate, a far greater
benefit of NFV is the potential for pervasive and compelling savings
on operating expenses achieved through increased efficiencies:
The
ability to share computing resources between functions

using hypervisor technology
Reduced
Use
power, space, cooling requirements
of widely available tools and skill-sets
Reduced
management costs through increased automation and

more efficient use of resources
Reduced
field upgrades (truck rolls) required than with

proprietary hardware
27
Virualization Benefits
Media Servers
and CDN
High
Automation Gain
Media Gateways
Network Apps
Packet Gateways
Radio Access
Network
Control Plane
Functions
Charging
Applications
Analytics
Platforms
NFV/SDN
Control
EMS/NMS
Edge
Routers
Core Routers and
Ethernet Switches
CPE
Network
Appliances
OSS/BSS
Automation and Optimization
Elastic scale
Resource pooling
Rapid deployment
Location optimization
Cost (TCO)
CAPEX
OPEX
Low
Less
Cost Gain
More
Source: ALU2
Virtualization stands to play a key role in helping operators predict

and manage costs as networks scale to accommodate exploding
demand for new services. According to the Infonetics survey,
operators see virtual enterprise CPE, or vE-CPE, as the top use
case for NFV opex benefits by virtue of increased service agility to
businesses. Mobile core/vEPC comes second, followed by service
chaining and a virtualized IMS core (vIMS core), all of which are of
particular interest to mobile operators.
2.3 Customers Benefit as Well

As operators simplify operations and potentially reduce cost,
benefits may extend to the end-user as well. NFV will help to
enable features such as self-service portals that allow customers to
change services in real-time. An attractive pay-as-you-go model
may also arise versus requiring users to purchase capacity for
future needs.
2 Network Functions Virtualizations: Challenges and Solutions, Alcatel-Lucent, 2013.
28
NFV Risks
and Challenges

Amidst exponential traffic growth and rising user expectations,
NFV essentially calls for the creation of a carrier grade cloud,
an unknown quantity to say the least. The biggest risks and
challenges inherent in virtualization center on guaranteeing quality
and performance.
Deploying latency-sensitive networking applications in the cloud may
impact performance in unforeseen or unacceptable ways, or limit
the ability to react to changes in usage profiles without incurring
degradation. Service level agreements (SLAs), QoS requirements,
and expectations for quality are already in place, and operators cant
afford to let them slip.
The same is true of reliability, availability, and security. Overall
performance and satisfaction must remain as good as they are
today, or get even better as virtualized environments scale in size,
scope, and complexity.
NFV standardization efforts continue, but more work, and proof, is
needed to fully inspire confidence. Operators are moving forward
anyway as weve seen, tackling strategic and technological
challenges in tandem.
The bottom line is: jobs are on the line with NFV, with huge
potential downsides both for moving too fast and not getting it
right, or moving too slow, and not getting it done.
30
3.1 Strategic Challenges

In embarking on virtualization, operators need to establish a
business case, and a plan for determining what to virtualize and
when. They then must determine how best to approach migration,
and establish a means for demonstrating it worked.
3.1.1 Challenge 1: What to Virtualize and When

Determining the value of NFV is essentially a game of evaluating
the tradeoffs between openness and performance, flexibility and
control, and quality and cost. At the highest level, operators may
also need to decide between an application-driven strategy where
they essentially test NFV out on a particular function, and a more
aggressive, platform-driven strategy designed to achieve substantial
virtualization much more quickly and cohesively.
Functions that are mostly control-plane-centric are great first
candidates for virtualization. In mobile networks, certain elements
in the EPC and IMS networks make ideal candidates to virtualize.
These include the MME, all Diameter servers (HSS, PCRF, OFCS
and OCS), and x-CSCFs (call session control functions). While
all of these have real-time performance requirements, they do
not perform any user-plane processing, and therefore dont have
the same criticality as devices such as SGWs and PGWs on the
user plane. CPE and broadband access devices are other strong
candidates.
In some cases, the benefits of virtualization are pretty clear. Other
elements are obviously less ideal candidates, such as those with
31
challenging real-time requirements. For example, the potential gains

achieved by virtualizing data-plane routers, Ethernet switches, layer
1 processing functions, and encryption may not stack up against the
cost, deployment effort, and potential performance tradeoffs. That
said, some packet processing applications are now running on more
standard compute platforms, and others moving to COTS if not yet
to fully virtualized platforms.
Between the two extremesobviously good candidates for NFV and
clearly bad candidatesmany networking functions and devices fall
somewhere in the middle, with a host of variables coming into play.
By way of example, elements with strict real-time requirements,
or that use custom ASICs, may be more challenging and require
more server blades, but still merit consideration due to long-term
efficiency gains.
For some functions, the benefits will be greater in certain
deployment scenarios than in others. In evaluating candidates
for NFV, a wide range of qualitative and quantitative criteria and
dependencies may come into play:
Can
services run together effectively in the same server?
If no substantial savings result from virtualizing a function, are

gains achieved through automation enough to make it a good
candidate for NFV anyway?
How
will it scale?
Scalability can be measured and achieved in many ways. Will a

single VNF be used to replace an appliance, or will many VNFs
be used or distributed? This can also impact service design.
32
3.1.2 Challenge 2: How to Go About It

Virtualization will typically occur in stages. This may delay the
ultimate benefits, but will afford operators the chance to evaluate
the risks and benefits during each phase.
NFV Application Evolution
Elastic/auto
scale
management
Network
cloud: app
on-boarding
SDN enablement
Features
Virtualized
appliance on
COTS/SHV
server
Migration to
COTS/SHV
server
Current
network
applications
Time
Source: HP1
When moving functions from physical appliances to virtual

instances, there can be many trade-offs and many design decisions
that need to be made:
How
should server resources like memory and CPU be

allocated to each function?
Should
resources be optimized for specific applications?
1 Technical White Paper: Network Functions Virtualization, HP, 2014
33
How
should NFV be integrated with what exists today?
Should
a contained or hybrid approach be used?
Where
will the VNF be deployed? Is there a regional data

center? Will the location of the VNF impact the service delivery
due to the network delay?
One challenge may arise from recent centralization of data centers.

For example, if a customer is located in California and the nearest
data center is in Colorado, operators must evaluate whether it
makes sense to provide a network function in the CO data center
and risk inducing latency.
Service providers must validate decisions and what if scenarios
each step of the way, asking other key questions such as:
How
much, how soon? The economic benefits of NFV will

likely increase as initiatives scale, but aggressive virtualization
means a profound transformation of existing paradigms. Most
operators will likely tackle the shift in phases, which means
having to define the various stages, and determine which
elements will be addressed within each, and in what order and
combination.
Who
should do what? NFV requires a combination of

extensive network and data center expertise. Operators must
decide:
Which members of which teams should be involved
Whether new or additional expertise is required
If and when it makes sense to outsource design, testing, or

integration
34
3.1.3 Challenge 3: Measuring Success

How do we know it worked?
In theory, the ability to meet SLAs and deliver reliable QoE to
subscribers shouldnt change when SDN and NFV are applied. And
once again, the flexibility achieved through virtualization should
ultimately result in quality changing for the better.
Measuring the success of virtualization is really the same as
measuring the performance of a dedicated hardware-based system.
The same KPIs apply with virtualization as before, and the two will
be compared against one another.
For NFV to be considered a success, the performance of the virtualized
system must be at least equal to that of the legacy system. And as
might be expected, problems may arise more frequently in the early
stages of adoption requiring different aspects of performance to be
evaluated as the various kinks get worked out.
Goals must be clearly defined and reliable strategies in place for
validating and improving performance from start to finish.
3.2 Architectural/Implementation Challenges

The primary architectural benefits of virtualizationelasticity,
nimbleness, and opennessrepresent challenges in and of
themselves. In achieving them, operators will need to dramatically
increase automation and real-time control of both physical and
virtual resources.
35
Management, security, and visibility strategies must also become

flexible and adaptable enough to address hybrid environments that
encompass both legacy and newly virtualized functions. High-level
architectural and performance challenges include:
3.2.1 New Devices, Techniques, and Dependencies

While the ultimate goal of NFV is simplicity, the process of
virtualizing new and traditional functionalities adds new degrees of
complexity:
Will
the IT model work? Can proven, reliable elements move

from custom hardware to unproven, and even untested
software-based solutions without performance suffering?
Will
combining particular functions on a single blade or

processor impact performance?
Will
bringing multiple elements like MMEs, SGWs, and IMS

core elements together degrade the performance of individual
components?
The adoption of NFV necessitates measuring the performance,

scale, and interdependencies of newly virtualized elements, as well
as the performance of the overall architecture end-to-end.
3.2.2 Speed
Will software be fast enough even with continuing advances in
CPU technology? The new Intel Xeon E5 v2 reportedly achieves
speeds up to 250Mpps, theoretically sufficient for most networking
36
applications, but providers need to be sure it all works in practice.

Having the right Network Interface Cards (NICs) within the server
also plays a role, especially while moving from 1G to 10G to 40G and
maintaining multiple NICs.
In addition to leveraging generic hardware as a platform, other
considerations will impact the performance of VNFs, mainly the fact
that there will likely be other VNFs running on the same hardware.
Multi-tenancy can introduce a degree of unpredictability into the
expected performance of each newly virtualized function.
3.2.3 Service Delivery

Virtualization should not impact existing services. The underlying
delivery mechanism, whether it be virtualized or not, should be
completely transparent to the services being delivered by the network.
This is critical since some network functions will likely be virtualized
in stages with other parts continuing to be implemented on a
traditional hardware-specific infrastructure. Throughout the
process, the availability of services should meet the same high
standards that are in place today.
3.2.3.1 Service Chaining
As noted above, one advantage of NFV is allowing more flexible
service chaining, or VNF forwarding graphs. In the traditional model,
most services had to follow the same path, whether there was
value in having them pass through specific functions (DPI, firewall,
IPS) or not.
37
With NFV, the operator has greater flexibility to define specific

service chains for specific services, such that a service only follows
the path that is directly applicable to it. However, in implementing
VNF FGs, operators must be able to guarantee sufficient capacity
and resiliency while increasingly automating provisioning.
These changes in the service chain should be tested prior to
deployment to ensure that changes do not impact service delivery in
unexpected ways.
3.2.4 Increasingly Distributed Infrastructure

Virtualized networks will become more geographically distributed
over time, making it harder to predict and control latency and
other variables. One of the benefits of NFV is the ability to quickly
instantiate systems, but sometimes the motivation behind this could
be the localization of services. This will increase the distributed
nature of the network itself, which can be taken as both an
opportunity (to reduce response latency times to the user) and a
challenge (latency times required to communicate with the home
system may need to be increased).
3.2.5 Multi-tenancy
Multi-tenancy becomes a challenge as the cloud-based approach
evolves. Operators must be able to manage policies for individual
services and flows as functions are decoupled from physical devices.
38
3.2.6 Scalability
NFV needs to be massively scalable to support large numbers of
data centers and millions of subscribers. In addition, the scalability
of resources will be far more dynamic in a virtualized environment.
The major advantage here is elasticity as VNFs can be created,
adjusted, and destroyed in real time, and on demand. When network
triggers are reached, capacity can be dynamically added or removed
from the overall network such that capacity and performance can
constantly change to reflect the current demand.
Networks must be capable of being reconfigured rapidly to
achieve the desired elasticity. Virtualized functions will need to be
dynamically updated as a result of scaling resources. For example,
the domain name system (DNS) service, which can be responsible
for load balancing between VNFs providing the same service, must
be instantly made aware of new elements being brought online as a
result of scaling.
3.2.7 Management and Orchestration

Operators must devise and implement new approaches for
managing virtual functions and networks. As part of this effort, 69%
of respondents in the Infonetics survey cited OSS/BSS concerns as
the biggest barrier to deploying NFV.
Operators need to figure out how to address virtualization across
complex back office operations and business support systems. Ideally,
the management of the virtualized elements will be transparent
relative to the non-virtualized elements. Thus it should not matter to
the management system if a function is virtualized or not.
39
3.2.8 Visibility
In a virtual environment, real-time visibility into the end-to-end
architecture becomes even more critical to guaranteeing service
availability and QoE. When things happen unexpectedly in either
the test lab or the live network, operators can leverage virtualized
taps (vTaps) and other monitoring tools to find bottlenecks, pinpoint
performance issues, and test varying configurations.
NFV and SDN will give rise to new monitoring challenges and
strategies requiring new visibility architectures and performance
metrics for components such as hypervisors and VMs, as well as
the performance of VNFs themselves.
Well take a closer look at evolving strategies for visibility in Chapter 7.
3.2.9 Robustness
Robustness is a broad challenge that refers to the ability of the
virtualized network to perform fault detection and invoke the
associated diagnosis and recovery mechanisms if a fault is detected.
This also means that the entire state of the faulty VNF must be
maintained as it is transferred and recovered elsewhere.
For example, if a particular VNF is found to be having problems, a
new instance of that VNF could be instantiated elsewhere in order
to replace it, but all the sessions within the first VNF would have to
be transferred to the new one. In some cases, it may be necessary
to fall back to a non-virtualized function from a faulty virtualized
function. This should be supported transparently.
40
3.2.10 Security
While virtualizing firewalls and load balancing functionality may
enable more flexible, nimble defenses, NFV also introduces new
elements, like hypervisors, that represent new targets for attackers.
The more nebulous boundaries of the cloud and increased
distribution of VMs across multiple geographic locations further
complicate defense strategies.
New configurations, devices, and defense strategies must be
validated prior to implementation and as networks evolve and scale
up and down.
3.2.11 Measuring Performance

A new approach to performance validation is required to fully
assess virtualization and the success of migrations end-to-end. The
next two chapters take a detailed look at evolving test architectures,
and real-world use cases.
41
Validation Strategies

Perhaps the greatest challenge operators face in undertaking NFV
and SDN is that getting started essentially means starting over.
Everything old, known, and proven, becomes new, unknown, and
unproven again.
In the process of becoming a VNF, tried-and-true networking
elements must be re-evaluated from the ground up with an
eye toward determining 1) Whether the business case supports
migration, and 2) How performance might be impacted.
Operators need reliable tools and strategies for quantifying the
benefits, costs, and risks associated with virtualizing individual
elements and systems, along with a means of determining whether
the process works and meets goals. New test capabilities and
strategies must be introduced, and the same rigorous performance
criteria applied.
Comprehensive NFV validation requires measuring the performance
of newly virtualized networking functions, as well as the ability to
assess the performance of the overall architecture and services
end-to-end. For the greatest efficiencies, a combination of traditional
and virtualized testing can be conducted, leveraging each approach
where it makes the most sense.
43
How Do You Get From Here to There?
Migration Phases
Start
Finish
Baseline
Confirm
Test
Assess
Plan. Test. Secure. Monitor. Optimize. Assess.
Test
Assess
4.1 Virtual Testing vs. Testing Virtualization

In tandem with server virtualization efforts, leading test solution
providers have introduced virtual versions or extensions of
traditional hardware-based capabilities.
4.1.1 Benefits of Virtual Testing

Virtual test solutions introduce several critical advantages such as:
The
ability to quickly recreate environments
Rapid
reproduction of development, Quality Assurance (QA),

and production environments through snapshots
Automated
Validation
creation and management of complete networks
of NFV configurations without truck rolls
Remote
debugging/reduced debug times
44
4.1.2 Which to Use When

The decision to use virtualized testing functionality versus traditional
hardware-based test systems depends on goals and requirements,
as well as the preference of the user. There are no clear-cut rules
here, but several considerations factor into the decision, and each
approach has its advantages for certain types and stages of testing:
Performance
testing. While virtualized testing appliances can

easily be used for performance testing, traditional hardwarebased systems have the advantage of known performance
expectations. These expectations are validated and
dimensioned under various conditions such that users know
exactly what theyre getting. In such cases, using traditional
systems simply removes a variable from the testing process.
Precise
measurements. When measurements with a high

degree of precision are required, typically with a resolution
of less than 1ms, then using dedicated hardware for the test
equipment is recommended. Specialized hardware will have an
advantage over generic servers for precise time-stamping of
the incoming and outgoing packets.
Development
testing. While bringing up a new service or

network element for the first time, operators may wish to run
single-shot functional tests. Here, the convenience of being
able to simply start a VM with a test tool instance can be very
powerful.
Resource
contention. When multi-person teams are

performing testing, having the ability to simply create your
own testing resource by instantiating it also proves powerful.
In this case, virtual test appliances can be quite beneficial.
45
Test
tool orchestration. Where there is a requirement for the

test tool to be managed and orchestrated by the same system
as the VNFs under test, using virtualized testing appliances can
provide a clear advantage.
Deployment
environment testing. This is a very unique

capability provided by a virtualized test tool, where the tool
itself can also be deployed in the cloud along with VNFs such
that tests can be run before deploying the virtualized functions.
Testing can be done post-deployment as well.
Visibility.
As networks will likely encompass both traditional

and virtualized network functions for some time, virtual
monitoring technology must also be added. Virtual taps
help pinpoint performance issues both in the lab and in the
field, demystifying the complexities and obscurity of newly
virtualized environments.
Physical Server
VM1
IxVM
Test Agent
VM2
IxVM
Test Agent
VM3
IxVM
Test Agent
VMn
IxVM
Test Agent
vSwitch
IxVM
Console
Ixias IxVM provides a software-based version of traditional hardware test ports
46
With so much occurring within a single server, it becomes

increasingly difficult to isolate the potential sources of performance
issues. This inability can result in finger-pointing and costly delays.
As well see in Chapter 7, a virtual visibility architecture is needed to
isolate issues to new elements such as the vSwitch, hypervisor, or
specific VNFs.
Standard Test Solutions
Virtualization Test Solution
Hardware-based traffic-generation / load

testing
Software-based traffic-generation / load

testing
Suitable for all performance, scale and

system tests
Ideal for development and functional

testing in virtualized environments
High-precision performance
measurements
Low-latency precision and accuracy.

Unlimited performance measurement
platform
4.1.3 Requirements for Both Approaches

It will become increasingly critical to leverage both traditional
and emerging virtual test capabilities. It will also be beneficial to
have the two platforms interoperate and deliver the same sets of
measurements.
4.2 Critical Components of Validation

Testing a virtualized implementation of the network requires
the same important capabilities used in testing a dedicated
implementation of the same network. Testing must be able to
validate, for example, that the EPC and IMS are capable of delivering
services with the same QoS, whether virtualized or not.
47
Assessing a virtual system requires measuring the various things

that can have an impact on performance. This includes assessing
the resources allocated to the VM (CPU, memory), ensuring against
overprovisioning, and allocating the best resources for the expected
performance. Critical capabilities should be validated as usual:
User-plane
performance (how much traffic can be carried through

the network without QoS degradation)
Capacity
Control-plane
End-to-end
Service
Critical
performance
QoS
validation
support features (charging and policy, DNS, DRA,
security, etc.)
4.3 Key Performance Indicators (KPIs)

In evaluating the general performance of mobile core functions, typical
KPIs include:
Total
user-plane throughput over default bearers (besteffort service)
Total
number of subscribers supported
Total
number of control-plane procedures per second supported

(includes procedures such as attach, detach, service request,
handover, IDLE mode to ACTIVE mode transitions, and the like)
Session
establishment availability and latency
Dedicated
bearer establishment availability and latency
Handover
latency
48
For QoE, related KPIs include:

For
best-effort, TCP-based services: time to first byte, time to

last byte, TCP retransmissions, TCP resets
For
conversational traffic, typically over dedicated bearers:

dropped packets, jitter, Mean Opinion Score (MOS)
For
VoLTE control plane: call establishment time, media

cut-through after answer
When dedicated bearers are used, other typical quality of service

(QoS) tests measure the performance of the guaranteed QoS
(dedicated bearer) traffic when the amount of best-effort traffic
is increased. In this case, the dedicated bearer traffic should
have a relatively unaffected QoS, even while the best-effort traffic
is increased.
Also see Ixias Authoritative Guide to Advanced LTE Testing for a
detailed discussion of mobile network test configurations and KPIs.
Many of the procedures and strategies outlined in that document
apply equally well to virtualized networks.
4.4 Scope of Validation

A basic premise here is that NFV can largely be considered an
implementation of the same functionality using new technology.
As weve said, the resulting system should at the very least deliver
equivalent functionality and performance as legacy systems that
use dedicated hardware. Knowing the baseline performance and
functionality of the legacy network and devices is a critical first step.
49
As such, the first and most important testing to conduct is

essentially a regression test to validate that existing features and
performance can still be guaranteed with the newly virtualized
implementation of the network.
However, many of the new technologies that enable NFV and its
various benefits have been proven effective for certain application
areas (such as cloud storage and computing), but are not yet proven
in real-time delivery networks such as the EPC and IMS core. As
such, the new technologies used in virtualization should be validated
specifically, in addition to the more traditional aspects listed above.
These new technologies, along with new concerns introduced by
virtualization, include:
Dynamic
scaling of capacity and performance: The elastic

properties with respect to performance are some of the most
valuable benefits of NFV.
Orchestration
and management of VNFs: While management

systems are not new, functionality is impacted by the fact that
functions are virtualized.
Fail-over
and resiliency: The ability to spawn new VNFs as

a result of detecting a fault in an existing one, as well as the
ability to fall back from virtualized to legacy functions.
Legacy
interworking: VNFs working in conjunction with

legacy elements, especially as it impacts existing functions
that have to work with both virtualized and non-virtualized
functions (for example, the RAN).
Portability:
Being able to move VNFs across multiple

hypervisors, hardware hosts, etc.
Flexible
service chaining: The ability to define different

network forwarding graphs on a service basis.
50
4.4.1 Managing the Migration

With NFV, theres no hard cut-over as with other infrastructure
migrations. At each phase, operators must begin with
baselining current functionality, then plan, test, and secure new
implementations. Upon migration, performance must be validated
again, then monitored on an ongoing basis to maintain quality.
Lab-to-live testing should encompass:
Planning:
Evaluating new devices and architectural models

on performance criteria vital to customer satisfaction (speed,
latency, reliability, overall application quality).
Design:
Replicating the complexities and variables of

virtualized wireless networks.
Quality:
Measuring/guaranteeing subscriber QoE for

critical applications in the face of interference, mobility,
outages, heavy traffic loads, and other variables. For
equipment manufacturers, this would also include
validating competitive claims.
Security:
Replicating security attacks.
Deployment:
Replicating field issues.
Visibility:
Maintaining QoE as architectures evolve and

services scale.
Regression testing is needed to ensure that existing functions are

not broken during the introduction of new variables (features,
hardware changes, hypervisor changes, etc.). This is a continuous
process that must extend beyond migration and start-up for the
lifetime of the network.
In the following chapters, well take a more detailed look at
evaluating components of the virtualized infrastructure, then at test
cases for real-world carrier network deployment scenarios.
51
Evaluating the NFV

Infrastructure
Chapter 5: Evaluating the

NFV Infrastructure
In moving forward with NFV, the virtualization infrastructure needs
to be selected carefully. Many considerations come into play and
each choice can significantly impact the overall strategy.
Common infrastructure elements will not only determine the
performance and features of the NFV system as a whole, but
potentially cause bottlenecks. For each element, specific aspects of
performance must be taken into account:
5.1 Hardware
Server features and performance characteristics will vary from
vendor to vendor. The obvious parameters are CPU brand and type,
memory amount, etc. Additionally, support for specific software
optimization APIs within the hardware can have a significant impact
on performance. The performance level of NICs can make or break
the entire system as well.
5.1.1 Driver-level Bottlenecks

At the server level, routine aspects such as CPU and memory read/
writes can cause underlying issues. And in most cases, more than
one type of server platform from more than one vendor will be in
play. Testing must be conducted to ensure consistent and predictable
performance across multiple platforms as VMs are deployed and
moved from one type of server to another.
53
Chapter 5: Evaluating the NFV Infrastructure
With physical NICs, performance can be impacted drastically by

simply not having the most recent interfaces or drivers.
Virtualization Bottlenecks
Virtual
Machine
Virtual
Machine
VNF
VNF
Communication Bottleneck
4 Virtual Machine Bottleneck
Host vs. Guest OS
Hypervisor
2 Virtual Switch Bottleneck
Virtual Switch
Driver Level Bottleneck
Server Platform
5.2 The vSwitch

There are many options and factors to consider in selecting a
vSwitch, some of which come packaged with the hypervisor,
while others are standalone. vSwitches vary from hypervisor to
hypervisor, with some favoring proprietary technology and others
leveraging open source.
Some hypervisor vendors have also designated specific functions

within this layer that others have not. For example, some vSwitches
provide very basic L2 bridge functionality while others act as fullblown virtual routers.
54
In comparing and evaluating options for unique production

environments, operators will need to weigh vSwitch performance,
or throughput, against resource utilization. Testing should begin by
baselining I/O performance, then progress to piling virtual functions
on top.
While provisioning the vSwitch, careful attention must be given to
resource allocation and the tuning of the system to accommodate
the intended workload (data plane, control plane, signaling).
Overprovisioning must also be avoided. The virtual equivalent of
throwing bandwidth at the problem, excessive allocation of CPU,
memory, and other resources not only causes waste but can
actually degrade switch performance in a virtualized environment.
The software used to implement virtual switching within the NFV
infrastructure is also important because the vSwitch introduces
new variables, and can become a bottleneck in the system.
5.3 The Hypervisor

Moving up the stack, operators again have multiple choices in
providers with both commercial and open source options available.
The commercial products may have more advanced features, while
the open source alternatives have the broader support of the NFV
community. And while it would be ideal to consolidate to a single type,
most real deployments will typically feature more than one flavor.
Hypervisors provide the ability to strictly provision virtual resources

(memory, CPU, and the like) to each VM. Most can loosely provision
them, enabling the ability to oversubscribe the server hardware.
55
They also provide the ability to start, stop, and snapshot a VM,
which enables backup and re-provisioning, or moving from the lab to
the network.
Hypervisors have common feature-sets, providing the ability to
virtualize the underlying server hardware and provision VMs, but they
also have unique features and performance. In selecting a provider,
its important to look at both the overall performance of each potential
hypervisor, and the requirements and impact of its unique feature set.
The ability of its underlying hardware layer (L1) to communicate with
upper layers should also be evaluated.
5.3.1 Hypervisor-level Bottlenecks

Hypervisors, along with the VM managers, provide the ability to move
a VM from one server to another (known as re-hosting) through a VM
migration. Performance during a VM migration will vary; some can
provide near-hitless migration while running. This portability feature
is a key requirement of NFV since VNFs will need to be moved if they
require more resources than the current server can provide, or if the
server needs to be taken out of service for maintenance. Hypervisors
will also vary in terms of their integration and support of automation and
orchestration software.
5.4 VM Manager
This component may or may not be part of the orchestration software.

One of the main decisions to be made is whether or not to use
OpenStack for the inbound API. OpenStack is becoming an industry
standard, and may facilitate selection of other elements, providing an
open and standard API (versus a closed system).
56
5.5 Management and Orchestration

Here, the profound fundamental shift from managing physical
boxes to managing virtualized functionality requires vastly increased
automation. Some operators are even investing in developing their own
orchestration system to handle provisioning and management.
Orchestration will be responsible for VM instantiation and
networking configuration for all VMs, which can become
critical in supporting the new benefits promised by NFV. Also,
the orchestration system will communicate with the element
management system of the overall network, and therefore will most
likely need to support the existing system deployed in the network.
The same considerations may come into play at this level as
for the VM manager. Is it an open or closed system? Are the
features supported by the system sufficient, and is it able to
control many types of elements via OpenStack? Is that important
to the overall strategy?
5.5.1 M&O Bottlenecks

VM managers are critical to keeping performance from becoming
bogged down. Communication with the orchestration layers directly
impacts the ability of the system to accommodate changes and
establish underlying networking connectivity on VMs themselves.
57
5.6 Virtual Machines (VMs)

Above the vSwitch and hypervisor, VMs themselves can impact
performance. Each requires virtualized resourcessuch as memory,
storage, and vNICsand each involves a certain number of I/O
interfaces.
In deploying a VM, it must be verified that the host OS is compatible
with the hypervisor. For each VNF, operators need to know which
hypervisors the VMs have been verified on, and assess the ability of the
host OS to talk to both virtual I/O and the physical layer.
5.6.1 Virtual Machine Bottlenecks

A virtual function on a VM will have its own limitations depending
on resource requirements and how applications are written. To
troubleshoot and ensure performance, the use of a virtualized tap
is critical. Virtualized monitoring solutions such as Ixias Phantom
vTaps monitor individual vNICs and connectivity from each individual
VM down to the hypervisor layer, helping to isolate and validate the
performance of the ultimate virtual application.
5.7 Real-world Scenarios

Once system requirements are understood, and parameters for
evaluating each component or function have been defined, test
methodologies can be easily repeated using either traditional or
virtual test solutions. After benchmarking and baselining small-,
medium-, and large-scale configurations, operators can easily redeploy snapshots of test scenarios allocating similar resources.
Lets take a closer look now at some specific test cases for common
migration scenarios.
58
NFV Test Cases

In this chapter, well take a detailed look at several use cases for
validating the virtualization of critical networks and functions. Well
consider three areas of operator networks that are prime targets for
virtualization:
The
Evolved Packet Core (EPC) in mobile networks
The
IP Multimedia Subsystem (IMS)
Broadband
access networks
Well also take a close look at how virtualized network elements work
in concert with legacy physical elements. The test cases described
here are performed in a lab setup designed to mirror an operators
production network. Individual VNFs can be evaluated as well as
multiple VNFs working together as a system under test (SUT).
6.1 Basic Test Setup

There are two basic test architectures for testing virtualized
functions:
The
first uses hardware-based test systems to emulate

subscribers, clients, and devices to test one or more VNFs.
The
second uses a virtualized test system. The virtual test

system acts as a VM/VNF within the same or a different
server as the VNF(s) under test.
Additionally, there will be instances where both physical and

virtual testers are used and both physical and virtual devices
are tested together.
60
6.1.1 Basic Test Setup Using a Physical Tester

The diagram below shows the setup for testing virtualized
networks using a traditional hardware-based solution. In this
example, one or more VNFs are provisioned on the server and the
network is provisioned to connect to the physical NICs of the server.
This enables testing with traditional equipment, in this case an
Ixia test platform.
Generic Test Setup for Testing a Virtualized
Network Using a Hardware-based Test Solution
VNF/Virtual
Appliance
VNF/Virtual
Appliance
VNF/Virtual
Appliance
vSwitch
Hypervisor
Generalized Server Hardware
Physical
Test System
NIC 0
61
NIC 1
Testing of mobile VNFs mapped to physical NIC interfaces

addresses many of the same aspects of performance as traditional
testing including:
Forwarding
performance (loss, latency, throughput)
Control-plane
performance (calls/sec, sessions/sec, maximum

concurrent calls/sessions)
Multi-protocol/multi-dimensional
testing
Unique testing and variables include:

Performance
of the vSwitch and VNF
Determining
the optimal resources (CPU/memory) allocated to

the virtual appliance to meet the performance requirement
Instantiation
of a service (how fast)
Termination
of a service
On-the-fly
Moving
changes to the performance of a service (elasticity)
(VM migration) a service
6.1.2 Basic Virtualized Test Setup

The figure below shows the use of virtualized testing at a high
level. Testing is inserted into the virtual environment using Ixias
IxVM virtualized test solution to provision test ports that are used
to evaluate data-plane or control-plane (protocol) functionality and
performance.
62

Network Using Virtualized Test Simulations
VNF/Virtual
Appliance
Virtual
Tester
VNF/Virtual
Appliance
Virtual
Tester
vSwitch
Hypervisor
Note: Simulations may or may not reside

on the same hardware platform (server)
Testing virtually by inserting test VMs into the virtualized server

includes:
Testing
the vSwitch for performance
Testing
each of the virtual appliances
Testing
the chaining of virtual appliances
Isolating
and testing each function before mapping to physical

interfaces
This scenario can be used to recreate various environments quickly:

Virtualization
technologies by their nature allow for snapshots
Development,
QA, and production environments can be quickly

recreated through snapshots
The
creation and management of complete mobile networks

can be automated
NFV
63
configurations can be validated without truck rolls
Testing can also be conducted with a combination of physical and

virtual test ports. Lets take a look at how these basic methodologies
and setups can be evolved to test realistic deployment scenarios.
6.2 Validating the Virtual EPC

The testing architecture shown below allows for the validation of
end-to-end EPC functionality. While the architecture shown uses
virtualized test components, the decision to do so can be made on a
case-by-case basis as discussed above.
In validating the virtual EPC, the regression testing described earlier
can be performed with high-load tests run to validate:
User-plane
performance: The packet-forwarding capabilities

of individual VNFs.
Control-plane
performance: The signaling capacity of the

network elements. Throughout the network, the control plane
can easily become a bottleneck.
QoS
and service validation: This is critical. Ultimately,

subscribers notice the quality of the services, and their main
point of reference will be the performance of the legacy
system. If degradation is noticed, it will be difficult to overcome
from a satisfaction standpoint.
Policy
and charging functions: Billing errors represent

another high-profile source of subscriber dissatisfaction that
must be avoided to prevent churn.
64
Testing the Virtual EPC

Virtual Tester
SUT
vEPC
vHSS
vPCRF
eNodeB
Internet
vSGW
vMME
vPGW
Emulated Subscribers
and Applications
vSwitch
Hypervisor
In addition, tests can be devised to validate NVF-specific

technologies and capabilities. These include:
Elasticity
of capacity and performance. By varying the load

amounts produced by the test system (sudden surges in
subscribers, user plane demand, etc.), the orchestration
system will be triggered to add or decrease performance and
capacity dynamically.
Obviously, this should be transparent to users and the services

delivered. Spikes in demand for capacity should be met
immediately, without causing any noticeable performance
degradation or introducing latency.
Resiliency.
By either generating or simulating faults in the SUT,

the ability of the system to diagnose and resolve those faults
should be verifiable. In such scenarios, sessions that are active
within the affected NFVs should be resumed as quickly as
possible, and new requests serviced quasi-immediately. The
65
amount of failed sessions or interrupted services should be

fully minimized.
Portability.
The same suite of tests run against VNFs using

a set of hypervisors and physical hardware can be run
against other hypervisors and hardware. This will validate
the portability of the VNFs, and also verify performance of
virtualized functions against various hardware platforms.
Multi-tenancy.
Different functions can be instantiated on

hardware platforms shared with other VNFs. For example,
one hardware platform could potentially run both a PGW and
a policy and charging rules function (PCRF). The ability of the
NFVs to perform predictably in this case, as well as the ability
to perform transparently, is verified (compared to each having
a separate hardware platform).
Also, see Ixias The Authoritative Guide to LTE Testing for a

detailed discussion of mobile network test configurations and KPIs.
Many of the procedures and strategies discussed in that document
apply equally well to virtualized networks.
6.3 Testing a Network with a Combined Legacy

and Virtualized EPC
This use case is important because of the simple reality that the
entire network will generally not be virtualized in one shot. There
are two basic cases where this can have an impact:
When
elements of the virtualized EPC will need to interact with

elements of the legacy EPC
When
outside elements will need to interact with both the

vEPC and the legacy EPC (prominent examples include the
RAN, the billing system, the IMS, roaming partner systems)
66
The figure below shows the basic test setup, using physical test
appliances. With the test equipment simulating both the RAN and
the IMS, as well as possibly a foreign EPC, the following test cases
should be attempted:
Handovers
between eNodeBs that are attached to the vEPC

and the legacy EPC
Inter-system
handovers between the local system and the
foreign EPC
Simulation
of a fault on a VNF, causing the session(s) to be

transferred to the legacy EPC as a fail-over mechanism
Combined Legacy and Virtualized EPC Testing
Physical EPC
Physical Tester
eNodeB
HSS
PCRF
Internet
Emulation of Subscribers
and Applications
MME
SGW
PGW
Physical
Virtual
Virtual EPC
HSS
MME
67
PCRF
SGW
PGW
6.4 Testing Virtualized Elements Within the EPC

As weve said, virtualization of the EPC will often occur
incrementally with control plane elements such as Diameter
servers, the HSS, and PCRF likely to be among the first migrated.
The test setup shown below can be used to validate the operation
of the vHSS (virtual Home Subscriber Server) and vPCRF,
which will be interacting with legacy EPC elements such as the
MME and PGW.
Two different architectures are shown:
vHSS and vPRCF Tested with a Core EPC
Simulation
Hardware-based
Test
TestingUsing
Virtualized
Elements within
theEquipment
EPC
Physical Tester
MME
PGW
SGW
Emulated EPC
Physical
Virtual
SUT
vHSS
vPCRF
Virtualized Functions
vHSS
and vPCRF Tested Using Virtual Simulations
Testing Individual VNFs by Simulating the EPC
SUT
vHSS
Virtual Tester
vPCRF
vMME
vSGW
vPGW
vSwitch
Hypervisor
68
Both test architectures aid in:

Understanding
performance (transactions/second) and scale

(number of subscribers, number of open transactions for the
PCRF) of the devices under test (DUTs)
Identifying
bottlenecks
Obtaining
accurate dimensioning information for use in

engineering the network
6.5 Testing IMS Virtualization

This use case focuses on validating the virtualization of IMS and the
associated applications, such as VoLTE. Like the EPC, IMS will not
be virtualized in a vacuum: the existing, legacy IMS and its services
will continue to exist for a long time to come, so interaction between
the two must be maintained.
Emphasis is placed on this aspect in the test architecture illustrated
below which includes:
Multiple
virtualized IMS networks and multiple application

servers (ASs) are placed under test. Typically, these will be
in different cloud environments, because IMS is likely to be
deployed in different geographic locations, creating a highly
distributed configuration.
real IMS, representing the legacy IMS that exists today,

along with its services.
simulation providing UE traffic for all applications, like VoLTE, file

and video share, and also new web-based applications such as
WebRTC. A hardware-based simulation is shown to illustrate the
distributed nature of the UEs, which, in a real environment, would
actually be connections to PGWs spread across multiple locations.
69
Combined Legacy and Virtualized IMS Testing

Physical Tester
Physical IMS
UEs
PSTN
IMS
Apps
Physical
Virtual
vIMS
Apps
IMS
The suite of tests to be run in this case should encompass the ability
to exercise and measure the following:
VoLTE
QoS: MOS (mean opinion score) measurements to

ensure voice quality is sufficient
Latency:
Call setup times, call connect times, application

response times
Capacity:
The maximum amount of calls sustained with

acceptable QoS
Testing should also include quickly ramping up the number of

simulated UEs, placing stress on the system in a short amount of
time to verify the dynamic capacity-increase functionality of the
vIMS. Operators must verify that capacity is downsized after the
number of UEs is reduced.
Demand on the applications hosted in the legacy network should
then be increased to verify capacity expansion into the vIMS.
70
6.6 Testing vBRAS Functionality

In the following example, a physical test system is connected to a
virtualized Broadband Remote Access Server or vBRAS, the device
under test. The tester emulates broadband access subscriber
traffic (e.g. homes with DSL connections) on one side and internet
services on the other. Important KPIs include subscriber capacity
and subscriber setup rates (PPP session rates).
Testing Virtualized Broadband Remote Access Server (BRAS)
Access
Subscribers
Physical Tester
Internet
Physical
SUT
vBRAS
vBRAS
Virtual
vBRAS
In this example, which is based on testing done by Ixia in conjunction

with 6Wind, several VMs were provisioned as vBRASs to achieve the
desired scale. This illustrates a major benefit of virtualization in that,
if the performance limit of a device is reached, additional VMs can be
deployed and the workload divided to achieve the desired scale
and elasticity.
6.7 Beyond Migration

Virtualizing network functionality is just the beginning. Ensuring
qualityand that the benefits of NFV and SDN continue to be
realizedrequires ongoing vigilance.
So lets take a look at the latest best practices in virtual visibility.
71
Maintaining Visibility
Chapter 7: Maintaining Visibility

Monitoring, Access, and Control in a
Virtualized Environment
Both in the preparations leading up to deployments and on deployed
networks, it is essential to maintain visibility into virtualized
environments as scope and complexity increase. Ongoing network
monitoring and visibility are also critical to security; specifically,
maintaining the effectiveness of tools such as intrusion detection/
prevention systems (IDS/IPS), data leak protection systems (DLPs),
and firewalls.
New and evolving visibility challenges include:
Pinpointing
performance issues. With so much occurring

within a single server, it becomes increasingly difficult to
detect whether problems stem from issues at the hypervisor
level, with the virtual switch, or from specific VMs. The inability
to source issues to specific networks or devices can easily
result in finger-pointing and wasted time.
The challenge can be complicated by hybrid administration
in which different teams or professionals have access and
visibility into different elements. Greater collaboration is
needed that can best be enabled through greater visibility.
Blind
spots. Gaps in visibility typically occur because traditional

security and performance monitoring tools cant see above the
vSwitch, where the east-west traffic flows. Inbound-outbound
traffic is visible, but only represents a fraction of the overall
visibility needed.
73
A loss of visibility into virtualized functions and traffic flow

between guest VMs can occur on shared virtualized hosts,
creating attractive hangouts for malicious intruders.
Performance issues may lurk here undetected as well.
Increased visibility is needed in order to detect and protect
what otherwise could not be seen.
The
momentum of virtualization. As weve seen, the benefits

of virtualization increase in proportion to how much and
how quickly the effort proceeds. As virtualized environments
expand and mature, it is vital that administrators have constant
access to reliable information on how performance data is
being used. Virtual machine sprawl may arise as VM creation
and cloning expands rapidly, making it harder to keep track,
and also to keep policies current.
With server sprawl, it is much easier to test and deploy QA

and development in staging environments. As these are not
productions systems, they dont have up-to-date security
policies in place, and thus present another gap in system
management.
Maintaining
full visibility into inter-VM or east-west

traffic within servers. Unseen inter-VM traffic on a shared
server constitutes a dangerous blind spot. In a traditional
environment, traffic is visible on the wire connected to the
monitoring tools of choice. Inter-VM traffic, however, is
managed by the hypervisors virtual switch without traversing
the physical wire that is visible to monitoring tools.
74
SPAN
and tap port limitations. Running in promiscuous mode

is the virtualization equivalent of physical network SPAN ports.
Operating in this mode degrades performance and opens up
new potential security breaches. Theres also no way to filter
specific traffica major issue in multi-tenancy environments.
Maintaining
compliance. As with physical networks, access

audit trails are necessary in the virtualized environment to
document compliance with industry regulations, and avoid
potential fines and bad publicity.
Load
balancing. Under virtualization, balancing loads efficiently

and separating network traffic becomes more challenging. VMs
move between physical servers, increasing the possibility of
untrusted VMs communicating with sensitive VMs on the same
virtual switch.
Complexity.
Multiple tools, probes, interfaces, processes,

functions, and servers are involved in virtualization. Monitoring
tools tasked with filtering data at rates for which they were not
designed can quickly become overburdened. Network speeds
are obviously dynamic (1/10/40/100Gbps), and without proper
filtering of traffic of interest, oversubscription and dropped
packets can occur.
As these and other factors keep actionable information from

reaching the right monitoring tools, migration schedules may suffer
along with performance.
75
7.1 Virtual Visibility: Evolving Goals

and Best Practices
To cope with the new and evolving challenges introduced by NFV,
visibility infrastructures must increasingly:
Operate
without negatively impacting the performance of the

virtual environment
Enable
regulatory enforcement across the converged physical

and virtualized infrastructures
Integrate
smoothly with virtualization technologies without

requiring architectural changes or adding a large footprint
Support
the elasticity of the infrastructure and follow

machines as they are migrated or vMotioned for optimized
performance
At the most fundamental level, the goal is to achieve visibility without

interferencea means of exporting traffic of interest from VMs
to the appropriate monitoring tools. Several options have been
proposedadding inspection VMs, installing clients on VMs to
capture and direct traffic, and otherswith each introducing its own
trade-offs, such as sacrificing VM or application-level vCPU, vRAM,
and/or storage.
As the complexities of virtualization unfold, neither traditional taps,
nor any other conventional solution appears to be able to capture
all the traffic that flows between VMs. Thus, a new virtualized tap is
emerging to bridge the gap.
76
7.2 Phantom Taps and the New Virtual

Visibility Framework
Ixia has defined a Virtual Visibility Framework that eliminates
blind spots, speeds application delivery, and promotes effective
troubleshooting for security, application performance, and SLA
fulfillment. Inter-VM traffic monitoring works to eliminate blind
spots, restoring visibility lost in virtualized server infrastructures.
Virtual Visibility Architecture
Hypervisor
VNF2
VNF3
Phantom vTap
vSwitch
VNF4
Blade Backbone
VNF1
Hypervisor
VNF5
VNF6
Phantom vTap
vSwitch
LAN
Network
Monitoring
Tools
77
A key part of Ixias comprehensive visibility architecture, the

Phantom Virtualization Tap (vTap) captures the east-west data
passing between VMs, and sends specific traffic of interest to
physical or virtual monitoring tools. A software-based solution,
the Phantom vTap deploys a module that resides in the hypervisor
kernel, passively monitoring all inter-VM traffic. It captures only traffic
of interest, without affecting production traffic.
Supporting best-of-breed hypervisors, Phantom vTaps eliminate
blind spots and improve performance by working to get packets
out of the virtual environment in real time with the least amount
of impact on the vSwitch. The vTap mirrors packets as they pass
between guest VMs with a minimum of overhead on the hypervisor
level and no significant processing performed (and in turn no
performance impact) on the virtual machine itself.
vTaps represent a major step towards mainstreaming virtualization,
enabling rigorous network management visibility and control
over sprawling virtual server infrastructures. Used with a
Network Packet Broker (NPB) switch, Phantom vTaps provide
the functionality achieved by traditional hardware taps and portmirroring technologies.
Though direct connection with an NPB, packets can be sent to any
existing security- or performance-monitoring tool. Smart filters can
then be applied to packet streams so that only data of interest is
sent to downstream management systems.
78
Ixia NFV Solutions
Chapter 8: Ixia NFV Solutions

Ixias virtualization test and visibility solutions encompass both
traditional physical test platforms and cutting-edge software-based
virtualized solutions. Overall, the solution is strategically architected
to cover all the bases as initiatives progress from the lab to live
deployments, and to introduce procedural efficiencies (like NFV and
SDN themselves).
Ixia virtualization testing assesses application and infrastructure
performance, as well as visibility and optimization throughout
the transition to the cloud. Advantages of this life-cycle approach
include:
Unified
test applications across both physical and virtual test

and visibility platforms
Software-based
A
traffic generation
virtual solution requiring no proprietary hardware
Consistent,
easy-to-use applications
8.1 Comprehensive Real and Virtual

Test Capabilities
As weve seen, carriers have two powerful options for testing the
performance of virtualized network functions:
Using
traditional hardware-based systems and mapping the

service to the physical server
Testing
80
virtually by inserting the testing into the virtual platform
Ixia delivers both, with each approach enabling measurement of the

same critical aspects of performance: forwarding rates, protocol
performance, scalability, etc.
Traditional and virtual test solutions support end-to-end virtual
validation of network, data center, and application performance.
Product offerings include traditional hardware-based IxNetwork
L2-3 and IxLoad L4-7 test solutions, along with IxVM, which
provides virtual, software-based versions of the two solutions.
8.1.1 IxNetwork
Service providers worldwide rely on IxNetwork to test routers,
switches, and other L 2-3 devices. Testing features high-load
traffic-generation at rates up 100GE, and supports a wide variety of
protocols including: IPv4/v6 routing, bridging, broadband, multicast,
MPLS, Carrier Ethernet, and SDN.
8.1.2 IxLoad
Ixias IxLoad L4-7 test solution emulates and validates the delivery
of voice, video and other application traffic, as well as malicious
traffic generated during security attacks. Delivering multiplay
service emulation in a single application, IxLoad provides ultra-high
performance and realism, including flexible subscriber-modeling in
evolving service provider networks.
81
8.1.3 IxVM
The virtual version of IxNetwork, called IxNetwork/VM, verifies
protocol functionality and validates SDN deployments. Similarly,
IxLoad/VM measures application performance in virtualized network
environments by providing stateful load testing of VM-based
services and I/O performance testing.
Network Using Virtualized Test Systems
Virtual
Tester
VNF/Virtual
Appliance
VNF/Virtual
Appliance
Virtual
Tester
vSwitch
Hypervisor
8.1.4 Phantom Virtual Taps (vTaps)

As described earlier, Ixias Phantom vTap provides a software-based
solution that supports all leading hypervisors (VMware vSphere,
Microsoft Hyper-V, Citrix XenServer) to deliver 100% visibility
into virtual traffic. Featuring integration at the hypervisor kernel
level, Phantom vTaps offer out-of-band monitoring that does not
affect production traffic, or require any services or agents to be
installed on VMs or at the application layer. The Phantom vTap is
vSwitch-agnostic and can mirror all traffic within the virtual switch,
apply smart filtering, and send only traffic of interest to specific
monitoring tools.
82
8.2 The Most Trusted Names in Networking

Trust Ixia
Service providers, NEMs, enterprises, and chip fabricators worldwide
rely on Ixia test solutions and assessment services to validate the
performance of devices, networks, services, and applications. Ixia
testing serves to speed time-to-market and validate the performance
of next-generation offerings, as well as the ultimate quality of the enduser experience.
Were working with carriers on front lines to:
Improve
network performance, resilience, security, and visibility
Speed
delivery of new services like VoLTE and richmedia services
Implement
and benefit from paradigm shifts such as HetNets,

SDN, and NFV
Ixia offers the industrys only end-to-end, lab-to-live solution for

ensuring the successful migration of critical networking functions
to software- and cloud-based infrastructures. With SDN and NFV,
everything old becomes new again, and well help ensure that it all
converges seamlesslybefore, during, and after virtualization.
83
Acronyms and Terms

AS:
Application Server
BRAS:
Broadband Remote Access Server
BSS:
Business Support System
CDN:
Content Distribution Network
COTS: Commercial-off-the-Shelf
CSCF:
Call Session Control Function
DNS:
Domain Name System
DPI:
Deep Packet Inspection
DRA:
Diameter Routing Agent
DUT:
Device Under Test
eNodeB:
Evolved Node B
EPC:
Evolved Packet Core
ETSI:
European Telecommunications Standards Institute
HSS:
Home Subscriber Server
IaaS:
Infrastructure as a Service
IMS:
IP Multimedia System
ISG:
Industry Specification Group
IT:
Information Technology
KPI:
Key Performance Indicator
LTE:
Long-term Evolution
M&O:
Management and Orchestration
MME:
Mobility Management Entity
MOS:
Mean Opinion Score
NF:
Network Function
84
Acronyms and Terms
NFV:
Network Functions Virtualization
NFVI:
Network Functions Virtualization Infrastructure
NIC:
Network Interface Card
OCS:
Online Charging System
OFCS:
Offline Charging System
OSS:
Operations Support System
PaaS:
Platform as a Service
PCRF:
Policy and Charging Rules Function
PDN:
Packet Data Network
PoC:
Proof of Concept
PSTN:
Public Switched Telephone Network
QoE:
Quality of Experience
QoS:
Quality of Service
RAN:
Radio Access Network
SaaS:
Software as a Service
SDN:
Software Defined Network
SDN:
Serving Data Network
SLA:
Service Level Agreement
S/P-GW:
Serving and Packet Data Networks Gateway
SUT:
System Under Test
TCO:
Total Cost of Ownership
VM:
Virtual Machine
VNF:
Virtual Network Function
VoLTE:
Voice Over LTE
x-CSCF:
Call Session Control Function
85
Acronyms and Terms
26601 W. Agoura Road Calabasas, CA 91302 USA

Tel +1.818.871.1800 | www.ixiacom.com
June 2014 915-3511-01 Rev. A

NFV Guide - Ixia

Uploaded by

Copyright:

Available Formats

NFV Guide - Ixia

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

NFV Guide - Ixia

Uploaded by

Copyright:

Available Formats

DEMYSTIFYING

A Denitive Guide to Successful Migrations

Copyright 2014 Ixia. All rights reserved.

Demystifying NFV in Carrier Networks

Introduction: Everything Known is Unknown Again

Chapter 1: What is NFV?

A Move Toward Software, Simplicity, and Standardization

1.2 The Market for NFV

1.3 Vendor Milestones

1.4 Carrier Milestones

1.5 NFV in the Mobile Core

1.6 Why Brave Another Paradigm Shift?

Chapter 2: Drivers for NFV

2.1 Increased Service Agility and Flexibility

2.2 Economic Advantages

2.3 Customers Benefit as Well

Chapter 3: NFV Risks and Challenges

3.1 Strategic Challenges

3.2 Architectural/Implementation Challenges

Chapter 4: Validation Strategies

Virtual Testing vs. Testing Virtualization

4.2 Critical Components of Validation

4.3 Key Performance Indicators (KPIs)

4.4 Scope of Validation

Demystifying NFV in Carrier Networks

Chapter 5: Validating the NFV Infrastructure

5.2 The vSwitch

5.3 The Hypervisor

5.5 Management and Orchestration

5.6 Virtual Machines (VMs)

5.7 Real-world Scenarios

Chapter 6: NFV Test Cases

6.1 Basic Test Setup

6.2 Validating the Virtual EPC

6.3 Testing a Network with a Combined Legacy and Virtualized EPC 66

6.5 Testing IMS Virtualization

6.6 Testing Virtualized Broadband Remote Access Server

6.7 Beyond Migration

Chapter 7: Maintaining Visibility: Monitoring, Access,

7.1 Virtual Visibility: Evolving Goals

7.2 Phantom Taps and the New Virtual

Chapter 8: Ixia NFV Test Solutions

8.1 Comprehensive Real and Virtual Test Capabilities

8.2 The Most Trusted Names in Networking Trust Ixia

Acronyms and Terms

Demystifying NFV In Carrier Networks

should they go about replacing their current network

will they know with certainty that the migration was

the virtualized solution deliver the performance,

With network usage, user expectations, and competitive threats

Demystifying NFV In Carrier Networks

with guaranteed quality. Ixia solutions validate that the motivations

Demystifying NFV In Carrier Networks

Why Read This Book?

NFV is and what changes

market dynamics and migration forecasts

best practices for evaluating and validating NFV

test cases for NFV validation

Well begin with a bit more background on what network