Network experience in the pocket

ComputerWorld Canada

Simon Fraser University (SFU) says it has tapped into a tool that takes an experienced engineer’s knowledge and turns it into a mathematical formula for troubleshooting different segments of the network.

Worth Johnson, the Burnaby, B.C.-based university’s director of computer operations and technical support, said appareNet, network intelligence software developed by Vancouver-based jaalaM technologies, has proven a useful tool to “help us solve problems quicker than any other tools we had seen or heard anyone talking about.”

The SFU campus network comprises 8,000 workstations and servers spread across 4,000 active ports on switches and hubs at 70 distribution points on three campuses. The network is linked to other educational institutions across the country through the shared fibre optic network; the province’s Internetworking society, BCNET; and CA*net, a national optical Internet research and education network.

Johnson said in the past, SFU’s network engineers had to use different pieces of test gear to check the performance of every segment of the network between the host or server end and the workstation. Then someone would be dispatched to check out what was happening in a particular segment.

“We could use tools like ping or traceroute to do testing, but those didn’t tell you anything over a particularly long period of time,” Johnson said. “If there were multiple segments along the way, if different protocols were being used, we didn’t know which segment was the problem, and we didn’t know the characteristics of the problem.”

According to Dr. Loki Jorgensen, director of research at jaalaM, a network engineer limited to working with a few ping tools or commercial network monitoring devices would have to “maintain a picture in their head of what the networks are like, where to find things and where to go to resolve problems and do expansions.” That’s not the easiest task, he said – it’s often a time consuming guessing game that only the most experienced engineers can successfully play.

Johnson agreed. “Most of us spend huge amounts of time dealing with the unknown area of what is really happening with the performance of network segments along the way.”

jaalaM’s approach is unique, Johnson said, because it “has taken a bunch of people’s experience troubleshooting networks, and has transferred that into a modelling process,” building it into an algorithm.

Jorgensen said appareNet builds up an end-to-end picture of network performance, and determines whether problems are application- or network-related, which problems are a priority, and when upgrades should be performed.

AppareNet uses network-sampling techniques to “send a few packets down the link lightly, in real-time,” along the same path travelled by the application. The packets are affected by delays, restrictions and limitations. When the packets return, appareNet uses proprietary analysis and mathematical modelling to report performance measurements.

The technology doesn’t require any remote agents. “Instead of plugging a device into the network and crossing your fingers, appareNet lets you see where the bottleneck or traffic is,” Jorgensen said.

Johnson says appareNet is “not cheap” and he wouldn’t recommend appareNet for community colleges or smaller universities with single campuses. But he said the investment is worth it for larger universities with multiple campuses, or with many campus buildings with various types of networking equipment.

The biggest benefit, Johnson said, is that any network engineer, regardless of experience, can use the tool and come up with the same results. There’s no real “rocket science” behind appareNet, but most technicians just don’t have the time to invent their own algorithms.