WHITEPAPER

VoIP Quality of Experience

Ensuring Your Network Supports Optimal VoIP Performance







More than any other type of application,
Voice over IP (VoIP) requires optimal
network performance to deliver an
acceptable quality of experience for users.
There are many reasons why VoIP is so
demanding. Unlike most data applications,
VoIP is sensitive to latency, jitter, and packet
loss. The maximum allowable latency of 150
milliseconds is quickly consumed by codecs,
serialization delays, router queuing, and by
transmission delay between locations.
Firewalls and network address translation add
latency to the processing of each packet.
Network elements induce jitter by
temporarily storing packets in buffers and
queues. Congestion or the misconfiguration
of Quality of Service (QoS) policies may
result in bursts of packet loss. In short, VoIP
is the most demanding application in your
portfolio requiring optimal network
performance, and the highest possible quality
of service levels to deliver an acceptable
quality of experience. As such, VoIP and its
related applications can present a
considerable challenge for those charged with
managing the performance of business-critical
data applications running on your converged
enterprise network.






Pg. 2 Copyright 2007 NetQoS, Inc. 877.835.9575
www netqos com
Converged Management for Your Converged Voice and Data Network
There are many performance monitoring tools on the market÷some help you monitor and manage network
performance, others focus only on VoIP applications. However, few provide the integrated metrics and visibility
required to deliver a superior VoIP quality of experience while ensuring network performance. By taking a
performance-first approach to managing both VoIP and your network using NetQoS products, you will be more
successful in deploying and supporting VoIP, while protecting overall network and application performance.

Network Performance Management Tools
Achieving your goals for network and application performance starts with the tools you select. Traffic analysis tools,
such as NetQoS ReporterAnalyzer
"
, leverage Cisco IOS
®
NetFlow data to give your network operations group insight
into the composition of traffic across major links, broken out by protocol and by class of service. These products allow
you to monitor for the proper assignment of Type of Service (ToS) settings to the traffic that is highest priority (which
should always include voice). They also enable you to monitor the unmarked, low priority traffic to establish whether
any high-priority VoIP traffic is mistakenly included in that class of service. And, they allow you to profile and monitor
the overall amount of VoIP traffic compared to other protocols on network links.

Some infrastructure monitoring products, such as NetQoS NetVoyant
®
, use SNMP to shed light on network
performance issues. In addition to providing data on network device health, SNMP is the access method for two
critical pieces of network performance information: Cisco IP SLA and Cisco Class-Based QoS (CBQOS).

As a built-in capability of Cisco routers and other devices, IP SLA sends synthetic network transactions from one
device to another, and returns data about the observed performance. IP SLA supports a variety of tests, including the
emulation of voice traffic for several different codecs and circumstances. While not a measure of actual production
traffic, it provides a consistent and repeatable characterization of the performance of any given path in the network.

CBQOS is a complex MIB which contains information about the performance of the class-based queuing mechanisms in
a Cisco router. With this information, you can verify whether or not packets are being dropped under load, and that
any packets dropped are from the lowest priority traffic queues, for example.

VoIP Quality of Experience Monitoring Tools
There are far fewer options when it comes to managing VoIP performance. Often, network managers must infer VoIP
performance from infrastructure performance metrics. However, monitoring VoIP call setup and call quality is critical;
monitoring VoIP performance without measuring actual call quality is like preparing a recipe without tasting the result.

And, with the increasing adoption and strategic value of VoIP applications, the need for VoIP performance management
is not likely to change. It has long been the quest of networking professionals to tie the network infrastructure clearly
to business issues to position themselves as an area of strategic importance. Until recently, network investments were
often seen as exotic forms of overhead, unrelated in any direct way to key business functions, like Order Entry or

Pg. 3 Copyright 2007 NetQoS, Inc. 877.835.9575
www netqos com
Sales. With the shift to VoIP, this changes. Use of the telephone is essential to business today; consequently, the
performance of VoIP instantly puts the network in the limelight. This brings a new layer of possible user complaints. In
many cases, it is also the first time the networking organization has primary responsibility for the performance of an
application.

Understanding VoIP Quality of Experience
¨Quallty o| e×perlence" ls a term tbat can be applleo to wbat your users are e×perlenclng wltb a vo|P telepbone
system. Wblle many aspects o| tbat e×perlence rely on your network's per|ormance, tbe subjectlve nature o| buman
use of the telephone warrants understanding these lssues as a ¨layer" ln tbelr own rlgbt.

Call setup is a good example. Long ago, in the old Bell System, the time it took to get a dial tone was quite long. This
has improved greatly over time and today, most users expect to hear the dial tone immediately after picking up the
handset. The typical standard for this delay is actually two seconds. If load, congestion, or even latency induced by
distance causes an IP PBX to respond more slowly, and the user does not hear the dial tone until four or five seconds
have elapsed, they will note this as a problem. Typically, users will hang up the phone after three seconds and try
again÷repeateoly. Ano, tbey'll report tbe quallty o| system per|ormance as poor÷even l| lt's actually worklng pretty
well after the brief delay.

The basic audio quality of a VoIP phone call is the other major dimension of VoIP quality of experience. As the joke
goes, the poor audio quality of cellular telephones prepared users for Voice over IP. In practice, people will quickly
notice and respond to audio problems on a telephone call. Long delays (like those experienced with old trans-Atlantic
calls) will cause people to revert to protocols familiar to walkie-talkle users (¨Over"), wblle notlceable ecbo levels
distract people from what they are trying to say.

Mean Opinion Score
To measure both subjective and objective aspects of VoIP audio quality, the Mean Opinion Score or MOS value is the
industry-standard measure of audio quality. Virtually all IP telephony equipment uses some form of quality
measurement which is mapped onto the MOS scale. The quality measurements take into account the underlying QoS
issues like packet loss and jitter, as well as the ability of the phone to compensate (with jitter buffering, for example).

These very specific, non-network-centric measurements for VoIP performance are the key drivers behind the need for
specialized VoIP performance monitoring products like the NetQoS
®
VoIP Monitor. NetQoS VoIP Monitor assesses
your users' quallty o| e×perlence and relates those measurements to network conditions to help you monitor VoIP
performance and troubleshoot the source of any problems. It provides comprehensive metrics reporting for MOS, call
setup performance (including time-to-dial-tone and time to connect), jitter, latency, and packet loss for every call, to
help you to better understand how VoIP is performing on your network. By using NetQoS VoIP Monitor with the
NetQoS
®
Performance Center product suite, you gain global visibility, via a single web-based management console, into
all of your applications÷data, video, and voice÷across your entire network infrastructure.

Pg. 4 Copyright 2007 NetQoS, Inc. 877.835.9575
www netqos com
Addressing the Social Challenges Associated with VoIP
In addition to many technical challenges, the deployment of VoIP is usually accompanied by organizational changes,
unoer tbe general beaolng o| ¨convergence." Punnlng volce tra||lc over an |P network brlngs togetber tbe concerns o|
these traditionally separate groups within IT. Telecom staff, used to dealing with the traditional PBX and phone
company issues, are suddenly working side-by-side with datacom staff, who speak the language of IP and routing.

This organizational change plays out differently in different organizations. In some organizations, convergence with IT
happens with a high degree of cooperation and success. In other organizations it happens with more finger-pointing.
One aspect of the problem is a lack of tools that relate the issues, traditions, and metrics of one group, to the issues,
knowledge, and metrics of tbe otber. |t's easy to see wby vo|P management presents multlple cballenges.

Making the Connection between VoIP Quality of Experience and Network Performance in the Real
World
With a performance-first management product that looks at the traditional telecom quality of experience metrics, like
MOS, and relates them to network-layer problems like packet loss and network paths, it is possible to manage VoIP
across technical and organizational boundaries.

Here's one use case scenarlo tbat lllustrates how VoIP quality of experience and network performance can be
connected. An engineer on the converged voice team of a large services organization is responsible for managing VoIP
performance. To quickly identify VoIP problems and pinpoint underlying network issues, he needs access to VoIP
application performance metrics such as call quality and call setup.

















Pg. 5 Copyright 2007 NetQoS, Inc. 877.835.9575
www netqos com
1) Using the NetQoS Performance Center, and the call quality data available from NetQoS VoIP Monitor, the network
engineer notes that call quality overall is slightly impaired and has actually been declining over the past hour.

By glancing at the lower left view which lists the worst performing locations from the complete set of locations in the
enterprise, Raleigh and Austin are the suspects for the locus of this problem.













Pg. 6 Copyright 2007 NetQoS, Inc. 877.835.9575
www netqos com
2) A click on the Raleigh link in the NetQoS Performance Center takes the engineer directly to the NetQoS VoIP
Monitor product, with Raleigh selected as the location of interest.
The NetQoS VoIP Monitor Performance Overview for Raleigh clearly shows some kind of degradation in the last hour.
The metrics summary, which is based on all calls with Raleigh as the location, points to jitter, buffer loss, and packet
loss. Austin is the only location with a corresponding quality issue, so, to narrow the search, the engineer clicks on the
Austin data.







Pg. 7 Copyright 2007 NetQoS, Inc. 877.835.9575
www netqos com
3) By selecting the Metric Detail view, the engineer sees that for Raleigh, it is apparent that packet loss rose in the last
hour, while MOS degraded, during conversations with people in Austin. (The Metric Detail page also includes many
additional statistics not shown here.)


Packet loss can arise from various sources. Queue discards are a possible source of packet loss, but VoIP traffic runs in
tbe ¨Platlnum" class o| servlce, so tbere sboulo almost never be any packet loss |rom tbe source. Nevertbeless, tbe
data warrants further investigation.

4) The network engineer returns to the NetQoS Performance Center to drill into traffic analysis for the interface
connecting Raleigh with Austin.

Looking at traffic grouped by class of service (ToS marking), the engineer sees that voice traffic is marked and reported
in the highest class of service (EF DSCP46 in the screen below).

However, packet loss could result from a misconfiguration of some kind, so he looks into the protocol distribution on
this link for the default class of service÷for traffic that receives no special quality of service handling.


Pg. 8 Copyright 2007 NetQoS, Inc. 877.835.9575
www netqos com


5) Looking into the default class of service on this Raleigh-Austin link, the Protocol Distribution for this class of service
shows that VoIP traffic recently appeared in this class.

This should never happen, since the latency requirement of VoIP traffic demands the best QoS the network can
produce. So the appearance of VoIP traffic in this class is the problem which requires corrective action.

Resolution
Unrelated maintenance or re-configuration of the Raleigh-Austin link one hour ago appears to have inadvertently
oroppeo vo|P tra||lc |rom tbe ¨Platlnum" class o| servlce lt requlres, to tbe ¨oe|ault" class o| servlce, resultlng ln packet
drops.


Pg. 9 Copyright 2007 NetQoS, Inc. 877.835.9575
www netqos com
The engineer restores the QoS policy on this interface to the correct values, and VoIP traffic is restored to the needed
priority, ending the call quality problem quickly.

Conclusion
The move to VoIP is accelerating with more and more organizations adopting VoIP each day. Telecom and datacom
sta|| are belng convergeo just as networks are belng convergeo. || lt basn't alreaoy oone so, vo|P may soon lmpact tbe
performance of your network and the requests made of your network operations group. Finding a way to connect all
the moving parts so that the network performance continues to perform well and support new business applications÷
telephony first among them÷is a challenge that requires visibility into new metrics and integration of your existing
network performance metrics. Organizations that can successfully connect VoIP quality of experience with network
quality of service will be successful in using and supporting VoIP.































Pg. 10 Copyright 2007 NetQoS, Inc. 877.835.9575
www netqos com
About NetQoS
NetQoS is the fastest growing network performance management products and services provider. NetQoS has
enableo bunoreos o| tbe worlo's largest organlzatlons to take a Per|ormance Flrst approacb to network management÷
the new vanguard in ensuring optimal application delivery across the WAN. By focusing on the performance of key
applications running over the network and identifying where there is opportunity for improvement, IT organizations
can make more informed infrastructure investments and resolve problems that impact the business. Today, NetQoS is
tbe only venoor tbat can provloe global vlslblllty |or tbe worlo's largest enterprlses lnto all key metrlcs necessary to
take a Performance First management approach. More information is available at www.netqos.com.









NetQoS Global Headquarters
5001 Plaza On The Lake
Austin, TX 78746
Phone: 512.407.9443
Toll-Free: 877.835.9575
Fax: 512.407.8629

NetQoS EMEA
1650 Arlington Business Park
Theale Reading, RG7 4SA
Phone: + 44 (0) 118 929 8032
Fax: + 44 (0) 118 929 8033

NetQoS APAC
NetQoS Singapore Representative Office
Level 21, Centennial Tower
3 Temasek Ave., Singapore 039190
Phone: + 65 6549 7476
Fax: + 65 6549 7001

Website: www.netqos.com
E-mail: sales@netqos.com

© 2001-2007 NetQoS, Inc. All rights reserved. NetQoS, the NetQoS logo, SuperAgent, and NetVoyant are registered trademarks of NetQoS, Inc. ReporterAnalyzer and
Allocate are trademarks of NetQoS, Inc. Other brands, product names and trademarks are property of their respective owners.
WP rev1 20070816