|
Occasionally, TMC Labs chooses to �spotlight� products
that cannot be covered using our traditional testing methods. These types of
products may be large carrier-class equipment, developmental software, or
embedded tools. In these cases, we do look at every option available to us,
including demos, white papers, comparative analysis, and other information
so that we could understand the product as close as possible to actually
testing it. After extensive research, we decided that Telchemy�s VQmon was
indeed worthy of mention. However, since we did not receive any VoIP end
points embedded with VQmon, we can not ascertain exactly how well the
product works, but the methodology is sound and comparison analysis supports
Telchemy�s claims of having quality means of testing for degradation in
voice calls.
VQmon is designed to monitor the quality of every call made
through a VoIP network, and operates on the received voice packet stream.
There are now two versions of VQmon: End Point (EP) and Stream Analyzer
(SA). The EP version is the first version to come out, and it was the one we
examined more thoroughly. It is embedded in any end point, such as an IP
phone or media gateway, where the end point is actually processing the voice
stream and therefore has a jitter buffer and codec and recognizes the notion
of ports. The SA version was just recently released. It is designed for
deployment in management devices such as firewall routers, probes, and
shapers. It uses a �jitter buffer simulator� in order to accurately
monitor call quality for these types of products. It does this by detecting
an accurate account of packet discards by the jitter buffer in the end
systems, which would normally report less than actually occurred. It also
allows consistent and corroborating metrics and algorithms from any point in
the network.
In some ways, VQmon is similar to that of the PSQM and PAMS
standards, which are means to measure the quality of VoIP calls. The main
difference is that PSQM and PAMS are based on average packet loss, where it
does not matter when or for how long the degradation occurs, only how often.
VQmon actually measures packet bursts instead of each individual packet so
that the Mean Opinion Score (MOS) can be affected severely if most of the
loss takes place during one portion of the call. Furthermore, the score also
depends on where in the phone call the degradation occurs. Learned from
extensive research of caller behavior, people are more annoyed with
degradation towards the end of a phone call than earlier. Of course, this
may partly because the degradation actually caused the caller to hang up.
Also, a person remembers the lack of voice quality better when it had just
happened. Makes sense, doesn�t it?
One last differentiating factor between VQmon and the other
measurement tools is that PSQM and PAMS requires that a pre-recorded speech
file be transmitted across an IP network so that the impaired file can be
compared with the original file, whereas VQmon does not require these test
speech files. Therefore, VQmon has the advantage of calculating voice
quality on live calls, which is especially useful to carriers and service
providers. From the comparisons we studied, VQmon usually tested better than
PSQM and PAMS for all of these reasons detailed above.
Besides studying the comparison data between PSQM, PAMS, and
VQmon, we also analyzed a software demo for VQmon EP. This demo allowed us
to select from a list of audio files and play the audio stream. When we
played the file, the R-Factor, which is a transmission quality rating that
is used to derive an estimated MOS, and audio output graphs showed us where
there was packet loss during the call. We could zoom into the audio output
graph to see more clearly about what happened when the call lost some of its
quality. The different colors indicated this: green indicated normal
packets, purple indicated dropped packets, blue indicated good quality
packets during a burst occurrence, and red indicated the shoddy packets. We
could also hear through our headphones how the packet loss effected the
sound quality. After the audio file completed, a network R-Factor, User
R-Factor, and overall MOS score were shown. There was also a button that
displayed additional statistics. To simulate another call with impairments
showing up at different times in the call, all we needed to do was press the
anvil-looking button. Over all, the demo did give us a sense of how VQmon
worked and how the MOS score is effected not only by packet loss (especially
when coming in bursts), but also by when the packet loss occurred.
With VQmon in place, we do think that the quality of service
would improve among service providers and networks using Internet telephony
equipment and services and can also give a better indication of when and
where a call is experiencing abnormally high packet loss.
[ Return
To The January 2002 Table Of Contents ]
|
SUBSCRIPTIONS
TMCnet LOGIN
WEBINARS
