Can the network carry the load?

When voice over IP first began appearing on business networks, many people expected to simply fire it up and start talking. But when they tried it, a lot of what they had to say never made it to the other end of the line – which was probably a good thing, because when they discovered how well VoIP worked on their existing networks, some of their language wasn’t very polite.Because many early VoIP adopters expected it to be plug and play, networks often got no pre-implementation testing to see if they were ready to handle voice, recalls Brad Masterson, product manager at Mississauga, Ont.-based Fluke Electronics Canada LP, a maker of network testing tools. That frequently led to problems when the IP telephony application was turned on.

In 2003 Jeremy Urwin, a technical sales support director at Telus Corp., told the story of LAN administrator who refused to have Telus check his network for VOIP readiness before going live, even when offered a discount on the work. “The first two calls went through flawlessly, “ Urwin reported. “The third call dropped, and guess who that was? That was the CIO.” Telus belatedly did the readiness report, the network was adjusted – and the customer got a new LAN administrator.

Such hubris on the part of network managers not only cost a few unfortunates their jobs, but battered the reputation of voice over IP technology and some of the vendors who sell it. As a result, some major VOIP vendors now insist on testing every network on which they install their wares to make sure beforehand that it’s ready for voice.

voice puts additional demands on the network

“From a customer satisfaction perspective it’s now become a requirement,” says Dave Zwicker, vice-president of marketing at Viola Networks, Inc., an Andover, Mass.-based maker of network management tools.

And so it should be. IP voice – and more recently, IP video – make demands on networks that more traditional applications don’t, and when the network can’t meet the requirements, the problems are very visible. “Most networks may not be able to handle those applications,” warns Frank Dzubeck, president of consulting firm Communications Network Architects Inc. in Washington, D.C.

And with VoIP, adds Alain Jansen, chief services architect at CNG Global Services Inc. in Ottawa, “you don’t have the luxury of picking up the phone and calling if the network doesn’t work right.”

In fact, it might be ideal if this idea of testing the network’s ability to handle a new application before going live with the new functions – as opposed to just turning it on and dealing with the problems as they arise – were to gain popularity outside the world of converged IP networks carrying voice and video alongside data. But don’t count on it. “In today’s day and age it just doesn’t work that way,” Dzubeck says. “You let it play and then you see how it plays and you go ahead and fix it later on.”

Tight budgets and schedules are part of the reason there isn’t more pre-qualification for new network applications, Dzubeck suggests, and divided responsibilities – one group is responsible for the applications and another is in charge of maintaining the network – also contribute.

However, network testing and troubleshooting tools are getting somewhat better at spotting signs of trouble before the help-desk phones start ringing off the hook with complaints from irate and frustrated users. And those capabilities will probably continue to improve – and be put to more widespread use – as the tools gain more capabilities and the people who use them become more comfortable with the idea of autonomics, in which machines are allowed to fix themselves rather than calling in human intervention every time something seems to be going wrong.

For instance, Vancouver-based Apparent Networks, Inc.’s AppareNet software is designed to spot network problems and performance trends, giving network managers the information to nip problems in the bud before end-users start to notice them. “No-one wants to stand around and watch the application fail and then respond to that,” says Loki Jorgensen, Apparent’s chief scientist. “That’s the reactive world.”

There are other tools that help with the particular network-management headaches that VoIP creates, such as Walnut Creek, Calif.-based WildPackets Inc.’s OmniAnalysis Platform, which recently added VoIP and IP video monitoring capabilities through an agreement with Telchemy, Inc., of Suwanee, Ga. Scott Haugdahl, WildPackets’ chief technology officer, says it can report on issues such as dropped packets and signaling problems that affect end users’ VoIP and video experiences, allowing network managers to see problems as they develop.

Haugdahl says WildPackets increasingly hears from customers who want to be able to spot problems before users complain.

Viola’s NetAlly addresses four phases of VoIP testing from network assessment through monitoring, management and troubleshooting. In its initial phase, NetAlly uses simulated VoIP traffic to predict how well a network will perform under various loads and where the problems are likely to arise, Zwicker says. The same software agents used for the initial assessment help monitor network performance once the voice application is up and running, he adds.

When the simulated calls identify problems, Zwicker says, NetAlly helps network managers drill down to find the specific network segments or routers where the problems originate, so they can fix them.

no single tool monitors every issue that can arise

CNG Global uses NetAlly, AppareNet and other tools in its business, which includes helping the customers of major VoIP vendors implement their systems as well as monitoring the performance of customers’ networks. CNG works with its customers’ network managers to assess what it will take to prepare their networks for VoIP and make the necessary changes. It takes a toolbox to deal with multi-vendor installations and all the issues that arise, Jansen says. “There’s really not one tool out there that seems to be able to capture all the different scenarios.”

Network testing tools are getting easier to use, Jansen says, but he would like to see a little more flexibility. He says CNG chose NetAlly because Viola appeared serious about addressing multi-vendor networks.

Tools like these utilize the Mean Opinion Score (MOS), a metric used in telephony to measure the over-all subjective impression of the quality of a telephone connection.

The MOS offers a useful handle on network performance from the end user’s perspective in a way that data network administrators haven’t been accustomed to. “It is fabulous insofar as it is a metric and we’ve never had application metrics before,” Jorgensen notes.

does the network satisfy the end user?

Inspired by the MOS, an industry group called the Apdex Alliance has set out to create a similar metric that works for more than just voice – a numerical measure of user satisfaction with enterprise applications. Haugdahl says WildPackets just recently started supporting Apdex, offering users of its software a way to see Apdex scores at a glance and use them to spot network degradation before users start to complain.

“They really haven’t had a way to do this before pro-actively,” he says.

It’s all very well to know if a network segment is working or receive an alert when something fails, Jorgensen says, but what administrators really need is a handle on what kind of experience end-users are having. “Essentially what we’re looking for is an ability to arm network administrators … with an ability to see what (performance) the network is providing,” he explains.

That’s a step toward the autonomic network, Jorgensen adds. There are four stages on that journey – the reactive stage, the pro-active stage, the automated stage and the ultimate nirvana of a fully autonomic environment. “Network management has essentially been stuck at step two.”

There are signs of progress, but the obstacle now may not be technological capabilities so much as it is human nature.

“The technology is now starting to be in place for the machines to talk with each other and fix themselves,” Dzubeck reports. The problem as he sees it is that the people aren’t ready: Many technicians are still not comfortable with automating the fine-tuning of the network. “I think it’s a generational thing,” he adds. Today, network managers tend to prefer that network monitoring tools report problems and perhaps recommend fixes, but wait for their go-ahead to perform the actions. “The next generation will just say ah, just let them go ahead and do it,” says Dzubeck.

Would you recommend this article?


Thanks for taking the time to let us know what you think of this article!
We'd love to hear your opinion about this or any other story you read in our publication.

Jim Love, Chief Content Officer, IT World Canada

Featured Download

Grant Buckler
Grant Buckler
Freelance journalist specializing in information technology, telecommunications, energy & clean tech. Theatre-lover & trainee hobby farmer.

Related Tech News

Get ITBusiness Delivered

Our experienced team of journalists brings you engaging content targeted to IT professionals and line-of-business executives delivered directly to your inbox.