[July 26, 2001]
Packet Voice Testing Tools: Don't Deploy
New Services Without Them
BY JOHN ANDERSON
For the past two years, global data traffic volumes have exceeded those
for voice. Meanwhile, according to 2000 estimates, global voice services
revenue is approximately $600 billion, while data service revenue stands
at about $65 billion. Although this discrepancy can be partially
attributed to the value of voice and its market demand, much of it can
also be attributed to the low cost structure of transporting data. It is
no wonder, then, that so many providers of voice services are
investigating ways to transport voice on data networks.
The combination of real-time voice services and data packet networking
results in new challenges that neither voice nor data networking
encountered as separate technologies. Multi-vendor interoperability, and
delivering acceptable quality for a time-sensitive service over a network
designed for data, are just two of the daunting challenges in deploying
packet voice networks.
The methods needed for testing and troubleshooting packet voice
networks in order to address these challenges are not provided in
traditional communications testing tools. This fact is evident in the
proliferation of new testing tools designed specifically for VoIP networks. But sorting
out the many different approaches, and determining precisely what needs to
be done to get a network up and running, can be confusing.
So how can a network operator make VoIP work while meeting quality
standards? The answer is to implement a comprehensive and carefully
planned testing program.
Defining Test Parameters
The first step is to define what parameters need to be tested and what
values need to be met. A packet voice network needs to be able to
establish a service event (e.g., connect a call), deliver an acceptable
level of voice quality, complete the service event (e.g., disconnect the
call) and create a record for the event. This means testing both the
operability and the quality performance of a network.
While basic call generators can test a network for overall operability
in connecting and completing calls, it is critical to also test call
signaling and control (S&C) in a network to determine if systems are
interoperating and where call failures are occurring. Testing this is a
necessary step for getting any packet voice network up and running.
Ongoing testing capabilities are also needed for maintaining and
troubleshooting the network.
Testing S&C protocols means testing the operation of individual
systems (call agents, soft switches, gatekeepers, gateways, IP phones) and
the operation of a network of systems, in accordance with the standards
being used.
Testing the quality performance of a network is not as simple as
testing its operability. A review of industry standards quickly exposes
the fact that, while some Quality of Service (QoS) standards exist, this
area is still lacking -- there are no encompassing standards to define
what overall QoS needs to be. Furthermore, no standards specifically
address QoS on a packet voice network nor specify how to balance QoS with
bandwidth reduction techniques like voice compression and silence
suppression.
However, a review of QoS testing methods indicates that there is a
common understanding of key parameters for packet voice networks. These
parameters include:
- Service establishment:
- Service reliability (e.g., call completion rates over extended
periods);
- Service availability (e.g., call completion rates during busy
hours); and
- Call set-up time or post-dial delay.
- Voice quality:
- Clarity;
- Delay;
- Echo (signal level and delay); and
- Signal loss.
- Packet performance:
- Packet loss, including erred packets and packets dropped from
jitter buffer;
- Packet latency; and
- Packet jitter.
Overview Of Testing Tools
The next and most important step in a testing program is the selection of
the right testing tools. The right testing tools can significantly reduce
the time needed for testing, and the overall time needed to get the
network up and running. In general, testing tools can be categorized into
test domains and test methods.
Primary test domains are:
- Operational testing, which addresses the functional operation of the
network in terms of interoperability among systems and the
establishment of calls and services. In other words, does the network
work?
- QoS testing, which addresses service performance and media quality.
How well does the network work?
Primary test methods are:
- Active testing, in which an external test system stimulates the
network (for example, by injecting traffic into the network) to
capture and analyze the network's behavior.
- Passive testing, in which an external test system unobtrusively
monitors the network's behavior in response to "natural"
stimulation (e.g., real customer traffic).
A rigorous testing program will encompass all four of these primary
test domain/method combinations for the purposes of network validation,
certification, monitoring and, most importantly, troubleshooting when
things go bad.
Types Of Testing Tools
Active QoS testing is performed with tools that can generate
calls over a variety of interfaces, and then measure the quality of the
media that is transmitted over bearer channels. Voice quality testers are
available that will generate calls on a network and measure various
parameters of voice quality. Call generators can also measure call
completion rates and set-up times. A voice quality tester can mean the
difference between months of trial-and-error attempts at meeting
acceptable voice quality levels, and hours of targeted design and
configuration changes that result in a consistently high level of voice
quality.
Some tools exist to emulate VoIP packet streams without generating
calls, but this method may not accurately measure voice quality,
especially if QoS mechanisms are inherent in the call set-up process.
Plus, this method cannot measure parameters such as call completion rates
and call setup times.
To actively test Real-Time Transport Protocol (RTP) packet performance
in a reliable manner, a tool needs to establish a "call" on the
network. Tools that can generate VoIP calls over a data interface and
measure loss and jitter are best suited for active testing.
Passive QoS testing is performed with tools that unobtrusively
monitor traffic and measure key QoS parameters. Some protocol analyzers
and network probes can monitor RTP streams and report packet throughput,
loss, and jitter. It is important to use a tool that can provide real-time
data capture and measurement at line rates. New tools are now emerging
that apply a bit of intelligence to these measurements to predict
subjective voice quality based on packet metrics.
While it is possible to get QoS metrics in a passive manner from
network elements themselves, such as gateways and switches, using an
external test system is by far the most reliable and accurate method. An
external test system will not be susceptible to failures and performance
degradations of the network under test. An external test system will also
provide a common and consistent method for testing throughout a
multi-vendor network.
Active operational testing is performed with tools that can
generate signaling and bearer traffic, and test the response of the
network in terms of functions performed. Signaling testers are available
to test call agents, softswitches, media gateways, signaling gateways, and,
in terms of signaling compliance, call control operations and even
signaling performance. Network emulators, such as gateway emulators, can
test the operation and performance of call agents and softswitches in
handling calls for large-scale network deployments.
Tools that can generate calls are also available for feature testing --
this includes testing call processing agents for features such as call
routing, call transfer and find me/follow me services. Feature testing is
also useful in testing interactive voice response (IVR) systems.
Passive operational testing is performed with tools that unobtrusively
monitor signaling, control, and media protocols; and capture data for
testing the operation of a network in establishing services. Protocol
analyzers rule this domain.
It is absolutely necessary, in both the deployment and operation of a
packet voice network, to be able to troubleshoot call setup. This includes
VoIP protocols such as SIP, H.323, and others.
Often, calls fail due to problems in layers 2-5. Troubleshooting VoIP
networks with a protocol analyzer can quickly uncover issues such as codec
mismatches, DHCP problems, and IP addressing problems.
Protocol analyzers offer network protocol decodes, analysis and call
trace capabilities that can expose within minutes a problem that otherwise
could take days to uncover by examining network equipment logs.
Not included in this description of passive operational testing are
operational support systems (OSS) designed for network management,
billing, traffic reporting, and fraud management.
Selecting Testing Tools
Selection of testing tools should focus on an overall platform solution
that comprises the different pieces needed. Testing solutions are now
available that can be centrally managed and remotely operated, and that
cover most types of testing. Distributed solutions are valuable for
testing a national or global network from a central network engineering or
operations center.
It is also important to identify the purpose for testing, and to get a
solution that is matched to that purpose. Some tools are geared for
development labs, other tools are best for troubleshooting networks, while
others are designed
for operational monitoring and SLA management.
Expert chefs claim the key to successful cooking is in the
ingredients. Similarly, the key to successful testing is in the tools that
are used. While there is an abundance of VoIP testing tools available
today, an effective testing program needs only a few tools to perform the
necessary testing operations. It is most effective for those tools to work
cohesively, and those tool vendors who can offer a coordinated set of
these tools provide the most value.
John Anderson is the IP telephony manager at Agilent
Technologies, Network Systems Test Division. Agilent is a global
technology leader in communications, electronics, life sciences and
healthcare. With 48,000 employees serving customers in more than 120
countries, Agilent had net revenue of $10.8 billion in fiscal year 2000.
Agilent's Network Systems Test Division provides telecom equipment
manufacturers, service providers and enterprises with a suite of network
testing solutions to test, monitor, measure, repair and troubleshoot PSTN,
LAN, WAN ATM, IP and 3G networks and products. To find more information
about Agilent's IP telephony testing solutions, visit www.agilent.com/comms/voicequality/.
|