When it’s time for change, it’s time for testing. Systems management specialists say keeping tabs on the network is a timeless commitment, but instigating change without thorough testing is nothing short of a reckless undertaking.
They say best practices entail constant monitoring of the network. To understand the flow of traffic, it’s critical to establish behavioural baselines. Knowing what’s normal can help predict a new application’s performance para-meters and how the network will respond to the change.
Robert Whiteley, a telecom and networks analyst for Cambridge, Mass.-based Forrester Research Inc., says monitoring should best be done at all times and testing, “when you really need to understand things in both a proactive and reactive way.”
Proactive tests are critical for any new application deployment, for whenever a key infrastructure device is added and for major location changes, says Whiteley.
“This is when you’re trying to get a better, real-time blueprint for your network. You can have your [systems management] wiring diagram, but it’s not going to show you what’s broken,” he said.
Whiteley says voice over IP is the canary in the coalmine for this — if you turn on VoIP in your network, it’s going to quickly reveal all your network’s flaws.
“If it doesn’t sound right, I won’t know why and that’s why I’ll need to test,” said Whiteley. “Maybe a certain link is too slow; maybe the quality-of-service mechanisms aren’t turned on correctly.
“Beyond the cable tests and the router and switch tests, you also have to look at quality-of-service tests. You could have set it up correctly and on the surface it’s operating just like you asked it to, but maybe you didn’t really ask it to do the right thing so it’s having a negative impact.”
What’s been typical in enterprises is that Ethernet switches have been installed with layer-3 capability, but are still functioning at layer 2 because the IP capability hasn’t been turned on, according to Rick Pearson, networking solutions marketing manager for Agilent Technologies Inc. in Colorado Springs.
“Before you start trying to run VoIP to your end customers, you’re going to want to test the layer-3 IP performance inside your enterprise under a variety of conditions,” said Pearson.
Because IP networks are so dynamic and voice so sensitive, the RTP traffic (real-time transfer protocol) must be monitored consistently for jitter, packet loss and latency, says Joe Haver, Agilent’s network systems business development manager.
According to Robert Hickey, an integration specialist at Panduit Corp.’s Canadian headquarters in Markham, Ont., it’s a good idea to test the performance of the cabling. “You want to make sure your patch cables measure up to the industry standard requirements for VoIP applications,” said Hickey. “So if your cable’s been installed a couple of years ago, for example, you’d run a channel link test, which will incorporate the patch cords you’re using to plug into the VoIP phones.”
For videoconferencing, says Hickey, “you’re going to need at least Cat. 6 (1Gbps transmission) performance out of the cable.”
The bandwidth demands for current technology and the need to comply with service level agreements compel a network manager to conduct testing, according to Darin Stahl, a research lead for London-Ont. based Info-Tech Research Group Inc.
Tests are required when completing moves, adds and changes to any network endpoint, says Stahl, and not just those attached to the backbone.
Stahl identifies multi-media, IP SANs and fibre-connected SANs (storage area networks), continuous data protection (CDP), power over Ethernet (PoE) and IP telephony as technologies that are stretching the capacity of existing cable plants.
“These bandwidth-guzzling technologies are forcing network managers to re-examine network design, testing tools and testing requirements,” said Stahl.
Brad Masterson, product manager for Mississauga, Ont.-based Fluke Networks Canada, says you need to create a baseline of your applications, your network usage and your server usage to give you an understanding of what’s normal for the network.
“Everything might seem fine and then some killer application like video conferencing and streaming kicks in. All of a sudden it changes how the network behaves,” said Masterson. “It affects voice quality [of the VoIP system], it could affect your call centres or it could affect your application response times.”
The network should be monitored, measured and tested before any application is rolled out, says Masterson, with pre-deployment in a lab environment to see what effects the application is going to have on the network. “Do pilot rollouts for certain groups or sections, and test each time you roll out a pilot,” he said.
Forrester’s Whiteley says the reason not a lot of reactive testing is done is because of monitoring. “In between those events that demand testing, just have some decent monitoring tools,” he said. “If you’re consistently monitoring, you shouldn’t have too many large-scale incidents that come up and surprise you.”
QuickLink: 055426
