To view the Report card. Although we have little control over many of the monkey wrenches that can be thrown into a network, some problems herald their development by sending out early-warning signs. For example, as applications are added, utilization often starts to creep up until it reaches a point where response time is impaired. Catching and correcting incidents like this before they become full-blown problems enables the network to run more smoothly, which, in turn, keeps users happier and frees network engineers to deal with other incipient problems as they arise.

Until recently, we thought our enterprise management platform could adequately handle the task of gathering information and reporting on the pertinent network statistics. After all, our vendor had

been telling us that the latest version would give us additional benefits, such as real-time monitoring and alerting, equipment configuration via a GUI and the ability to document our network--all from one centralized console. But after investing almost two years attempting to prove that this was more than marketing hype, we acknowledged that though enterprise management platforms do a lot of things, they don't do any of them very well, including proactive network management in a multivendor environment.

Network Computing decided the time had come to look closely at products that perform trend analysis and reporting. Three of these--Concord Communications' Network Health, DeskTalk Systems' TREND and Kaspia Systems' Network Audit Technology--claimed to be able to pull relevant SNMP data from our multivendor network and provide reports that would alert us to developing problems and trends. After running them on our network for a couple of months, we concluded that these products accomplished this task much more effectively than our enterprise management platform had ever done, and with a lot less pain on our part. However, we also discovered that despite doing a good job of collecting and reporting data, these products would benefit from a fair amount of additional expertise and manual labor. In fact, it's difficult to see how any of these products could be implemented successfully without having someone on staff who possesses a solid understanding of the mechanics of SNMP.

Of the three, Concord's Network Health won us over. It was the easiest to use and sported a number of built-in features that let us get most of the data we needed, when we needed it. In contrast, DeskTalk's TREND was very difficult to use but included great development tools for those with enough time and expertise to develop their own reports. Finally, though Kaspia's Network Audit surpassed the others in terms of being easiest to get up and running, it was not as flexible as the other two. For example, it offered only one polling interval--and the only way to change this preset, recommended 15-minute interval was to edit Windows NT's Registry.

One area that caused a lot of difficulty was the way SNMP identifies the actual ports and interfaces on a device, usually through an index of sequential numbers corresponding to each port. Unfortunately, a port labeled "1" on the front of a switch is not necessarily the port that the index labels as "1," for example. Many times, the reporting applications would refer to an interface index number, which would force us into the tedium of manual labor--querying a table of interface index numbers and descriptions, then reconciling them with the physical descriptions by hand. Exacerbating this confusion is the likelihood that all the numbers can change on modular devices if an interface is added.