Service providers and enterprises know that if they want to provide VoIP they must find out if their networks can offer the service at a quality level that is close to the standards of circuit-switched telephony.
As a result, new testing tools and services are emerging to provide such an examination.
In April, NetIQ Corp. (www.netiq.com) released a VoIP module in conjunction with the upgrade of its Chariot testing tool. Chariot emulates transaction traffic from real applications and primarily evaluates the effect of changes to the network before they are rolled out.
Chariot uses end points and scripts. End points are software agents that reside throughout the network, generating traffic between themselves and measuring things like response time and throughput. Scripts define what kind of traffic those end points generate.
"What we've done with this voice over IP module is we've taken it one step further in the voice over IP area by allowing people to not only evaluate the voice over IP, but to evaluate the impact of voice over IP before you roll out any kind of voice equipment into the network," says Kim Shorb, Chariot product manager.
Service providers that roll out VoIP need to understand whether the voice traffic is going to be handled correctly, Shorb says. That doesn't necessarily mean the load that VoIP traffic may add. It could mean all of VoIP's unique characteristics.
"Do I have small enough lost data that it's not going to be noticed by the user, and how much delay do I have in the network?" Shorb explains. "Those types of characteristics can really cause voice over IP traffic to be degraded to the point where it would not be acceptable to the end user."
A second piece of information is whether rolling out VoIP will affect the applications the service provider uses.
"If I've got order-entry systems or something like that that basically run my business, I want to make sure I don't allocate so much bandwidth to VoIP that all my other applications go to pot, so to speak," Shorb says.
The Chariot VoIP module generates voice traffic between the modules which, to the network, looks like normal VoIP traffic. Then it generates an overall voice-quality score, called a mean opinion score (MOS). If the score isn't what's expected or shows the voice isn't at the required quality, then statistics can be gathered on such characteristics as one-way delay and jitter to determine where the problems are.
"What we do is we provide all these statistics so people can take that information and go fix the network," Shorb says.
Beyond the Score
Providing this data is important be- cause a mean opinion score doesn't tell engineers enough.
"If you can go to them with a thing that says my estimated MOS is 2.5 and, by the way, I also have X amount of jitter and X amount of delay and lost data between these two network segments, then the network guy has something that he can go and use and troubleshoot and fix the network before you ever deploy the voice over IP application into that environment," Shorb says.
For example, a bank with multiple branches may look to use VoIP in its network connecting those branches. Chariot end points can be established at each branch, and with its VoIP module, tests can be run remotely from the bank's headquarters to determine the voice quality between headquarters and each branch and between the branches themselves. The tests also could determine how quality is affected as network call load increases.
"I can see where things start to degrade so that I know what my network can accept today and what changes I might need to make to the network before I roll it out full force into the network," Shorb says.
The network's ability to handle VoIP is a factor on whether deployment of the technology will be successful, Shorb says, explaining that "voice calls are so sensitive to things like delay and lost packets, and things like that that are not typically a problem with a standard or typical IP application."
Shorb cites Avaya Inc. (www.avaya.com) CEO Donald K. Peterson's statement that 85 percent of today's router-based networks are not ready for voice deployments.
Avaya offers a service for predeployment testing of the networks' ability to handle VoIP. The service also uses NetIQ's Chariot test product as a key component.
"The only way today to inject actual traffic onto a network is either to use the actual product, or find some sort of application that can, in a controlled manner, simulate it," says Kenneth Kane, Avaya senior technical manager.
Earlier this year, Avaya introduced its basic network readiness assessment offer.
Customers provide information about their networks, such as its topology and what vendors and products used it. An assessment of a single building with no more than 400 users costs $1,000. For single-building networks with more than 400 users, the cost is $2,000.
Kane says Avaya evaluated several tools before settling on Chariot.
"The trouble with a lot of tools is that they normalize or average in the results of those traffic monitorings," Kane says. "So, for example, if I've had one in 10 phone calls from a department across the data network, one in 10 are bad. If you normalize the data over a 15-minute or one-hour interval, you have pretty much erased any ability to identify that there was a problem because, if you look at the average results, you've masked that one call that was the problem, you can't find it."
Avaya uses its own proprietary tools and Chariot. They allow VoIP traffic to be injected and monitored every 10 seconds.
"That gives us the ability to identify even a single call, even a small time period of a call when there is a problem, and then track that in order for us to then go back and diagnose and, hopefully, come up with a resolution of the problem," Kane says.
It's not unusual to find that a customer's network is woefully inadequate for the unique requirements of VoIP, Kane adds. Customers that have network infrastructure based on technology from the mid-1990s can't support things like virtual local area networks.
"But, in every instance where we dealt with those customers, they were very open to us documenting this so they could go to the decision-makers in the organization to get funding to upgrade their environment," Kane says. "So, in each of those cases, they openly embraced the results of our reports."
NetIQ's Shorb says an immediate and growing demand exists for the pre-employment VoIP testing her company and Avaya offer.
"There's a lot of interest (in VoIP) and, I think, there's a lot of hesitation," Shorb says. "People are very hesitant to roll it out because of the concerns [about overall voice quality]. It's the most business-critical application you could ever get because, if someone can't make a phone call in your network, the business comes to a halt very quickly."
Compounding the problem is VoIP newness.
"So, the easier you can make it for the person who knows the network very well to also be able to quickly determine the quality or how that network can also support new applications like voice over IP, the more valuable a tool is going to be," Shorb says.
Post-deployment Testing
However, once VoIP is deployed, other tools will have to be used to test the network and to ensure voice quality remains at an appropriate level.
Companies such as Empirix (www. empirix.com) perform load and performance tests using actual voice scripts.
A common Empirix VoIP test begins with a simple call--to see if it goes through--then ramps up the number of calls to find the breaking point, says Tom Lynch, who manages Empirix's VoIP development group.
The scripts accurately represent the sort of voice traffic the network will handle. That means they must include several voices because each voice has different qualities.
"We went to great pains to get voice recorded professionally and we send a variety of voice clips with different frequencies that exercise a voice range so we can make sure the gateway is performing adequately over that range," Lynch says. "So, it's important to use real voice from that perspective. We do the same thing with packets. We'll test with packets of real voice."
And while standard algorithms such as perceptual speech quality measurement (PSQM) and the perceptual analysis measurement system (PAMS) can measure voice quality, Lynch says that metrics can lie.
That's why VoIP testing professionals say that sometimes the most valuable tools are the ones they were born with: their ears.
