The Board of Directors put you in charge of the VoIP project. Bandwidth,
billing/cost, security, and voice quality are their main concerns.
Unfortunately, like many CTOs from the enterprise to the ITSP, you have no
idea how to ensure those issues are dealt with. Without doing so, you can't
begin to address enhanced services, but you might begin addressing your
resume.
Most telephony engineers would tell you that the answer is possession of
a good quality of service (QoS) tool, such as NeTrueQoS 2.0. We tested this
product in our laboratory against both real traffic from Microsoft
NetMeeting and simulated traffic/network degradation from Shunra Software's
The Cloud. When it came time to finish the testing and write this review, we
smiled. Good software makes us happy, and our fictional hero not only kept
his job, he got a raise.
INSTALLATION
From a bird's-eye view, it's easy to understand how NeTrueQoS works. Imagine
that there are three entities: One is your gateway in Los Angeles, one is
your gateway in New Delhi, and one is you, a long time ago, in a galaxy far,
far away. NeTrue's manager program resides at your location; and NeTrue's
agent applications are in California and India. Somewhere else in time and
space your SQL 7.0 server resides, and it's all connected by your choice of
a public or private intergalactic TCP/IP network.
Our publisher wouldn't sign the paperwork for a TMC Labs spaceship
(perhaps he's hoping we find our own ride out of the galaxy), so we were
forced to emulate those conditions here in the office. We configured three
Windows NT 4.0 servers on a network. We designated one of them as
"machine A" and on it we installed SQL 7.0 plus the NeTrue manager
and agent applications. We also configured this machine with two network
cards to make an ersatz router, because the product is intended for use on a
WAN, not a LAN. Our intention was to access the test from machine A (home
base) but actually run the test from machine B (Los Angeles) to machine C
(New Delhi). As the NeTrue engineers explained to us, in this configuration,
only the PoP at machine B needed the agent software. Machine C only needed
to stay put. NeTrue's minimum requirements for these computers are a Pentium
II 400 with 128 MB of RAM and eight GB of free disk space. Also, it must be
running Windows NT 4.0 Server, the SNMP service, and service pack 5, with
static IP addresses. SQL 7.0 can run anywhere on the network, but take
special care not to put it somewhere where network performance would suffer.
Once this instant WAN was configured, we ran the requisite installation
wizards. Part of this process involved determining the size of the SQL
database to be created, but otherwise, conducting this part of the
installation was very simple. After the reboots, everything was ready to go.
DOCUMENTATION
NeTrueQoS includes three documents. One is for installation; one is a user's
guide; and the other is a white paper about VoIP testing. We feel that the
installation guide can use more detail and better descriptions of sample
applications. The user's guide is fine, with good explanations of all of the
features. The white paper is superb, and should be considered required
reading for anyone who wants to learn about the Internet telephony industry
(toward that end, it's available as a download from NeTrue's Web site). The
bad news is that the program has no online help -- technically, there is a
help option in the GUI's pull-down menus, but it's only an "About
NeTrueQoS" screen with the version number and copyright information. As
far as TMC Labs is concerned, online help in a Windows application should
not be an optional thing.
FEATURES
This product's main features are an SNMP interface, real-time and scheduled
monitoring, and advanced reporting. Once network presences are added to the
map view(s) using standard SNMP conventions, double-clicking on the location
ID accesses location settings. This brings up a GUI with seven tabs. They
are LAN general configuration, destination IP configuration, current agent
task, scheduled tasks, real-time QoS monitoring, real-time trace route, and
historical data analysis/reporting. Of these menus, the first three are
quite simple to learn, while the latter four are more complicated.
The LAN general configuration menu includes the settings from the initial
SNMP session -- which are the LAN ID, LAN name, router IP, subnet mask,
agent IP, agent community, country code, country name, network uplink (in
BPS), and (X, Y) coordinates on the country maps plus information on the
status of the agent, mediator, and database maintenance applications. From
here, you can also get and set information to the agent, and you can verify
the system status. Next, the destination IP screen lets you configure both
network-side and trace route thresholds for the throughput, latency, jitter,
packet loss, voice index, and MOS (mean opinion score) of each targeted PoP.
This is one of the product's most intuitive, valuable features. Then, the
current agent task simply indicates the destination IP address, beginning
date and time, interval(s), packet size, packets per sample (per testing
frame), and timeout setting.
But as we said, the next few options require more patience to master.
"Scheduled tasks" is the most complex option. Using this feature,
you can preconfigure network-side or trace route QoS testing to run at any
time. For example, for longer tests, you can configure the schedule to run
during a weekend, as we did (see the Operational Testing section below).
Testing results (both real-time and scheduled) are managed using the
historical data analysis and reporting menu . Reports can be generated by
time intervals, Top-N, and among multiple destinations. You can also control
which kinds of records are purged first when the database becomes full. The
real-time trace route menu does what its name implies, with the option of
viewing routes across multiple network hops. There is the real-time QoS
menu, which uses colorful line graphs to illustrate the real-time
information as compared to the settings of the Destination IP menu (maximum
throughput, latency, jitter, packet loss, voice index, and MOS). There's
also a quick menu in this interface for setting the destination IP address,
packet size, packets per sample, timeout, and refresh interval, plus a
start/stop button.
OPERATIONAL TESTING
We used the schedule feature to conduct a 24-hour QoS check on our three-PC
WAN. To save resources in the laboratory, we had the SQL server, the QoS
manager, and a copy of the QoS agent all on one computer. By connecting this
computer across the dual-NIC'd machine to the third computer (second agent),
we simulated how a real-world QoS setup might work. We used Shunra's The
Cloud to customize degraded network conditions. Combined with the threshold
settings of the NeTrueQoS, the kinds of testing scenarios we could create
were limited only by our imaginations.
With such an advanced testing tool, it's important that the end user
understands how to read the results. Some of the concepts are easier than
others. Consider these IP telephony truths:
- Increased bandwidth will always improve performance.
- Other applications sharing the network will hurt performance.
- Audio quality (MOS) is correctable with echo cancellation, jitter
buffers, and redundant packeting; but improving latency by using higher
compression rates has the opposite consequence.
- Good QoS monitoring is needed throughout the network, to ensure that
time-critical voice and video packets are prioritized ahead of data and
fax packets.
Considering its power, we found that this product is relatively easy to
learn and to use, especially compared to its competition. Other solutions
are available for bit- and memory-level analysis, and for protocol analysis
of H.323, MGCP, SIP, etc. But for a Windows product that operates parallel
to (rather than with) your network traffic, this product impressed us.
ROOM FOR IMPROVEMENT
Of the alternative products mentioned in the preceding paragraph, we feel
that H.323 testing is the next natural step. It would be extremely useful if
the main GUI had a resizable window. Also, this product is designed
exclusively for WANs and wider environments, but considering that many
enterprises implement Internet telephony applications on local networks
today, we would like to see a LAN version. Finally, conspicuous online help
is another issue: In Microsoft Office, the animated assistants are so
conspicuous that they are annoying. This product goes to the opposite
extreme: The "online help" is conspicuous because there isn't any.
This is unacceptable in all but the simplest products.
CONCLUSION
The NeTrue engineers created a product that is extremely usable despite some
serious flaws, but we're not sure if this should impress us. Regardless, the
product is powerful, reasonably priced, and simple to maintain. In
conjunction with products like Hammer Technologies' Hammer.323 and some of
the lower-level network testing devices, we feel that the NeTrueQOS 2.0 is a
good value and will serve you well. We think very highly of this product, we
will go on the record as recommending it, and we might even use it in our
laboratory for testing other products. However, until some of its issues are
addressed, we are reserving our Editors' Choice award for a future version.
|