Monitoring the Big Picture

I installed the software on a Solaris box in our Network Computing Real-World Labs® at Syracuse University to monitor devices on SU's production network. The setup took about three hours, including the OS, patches and software. The improved documentation and install procedures show the application's maturity, with cookbook-style directions that even Unix beginners can perform.

Accessing the Network

You can add network devices, systems and applications using the browser user interface or the command line. In fact, you can use the command line for nearly all tasks and script writing--with the exception of viewing graphical reports. Access to the ProactiveNet server is through telnet, SSL or HTTP Tunnel Proxy.

It's a snap to make changes to all server monitors. You need only go to the Java administration application in the user interface and right-click on a device. This brings up a list of actions that can be launched for the device. Most of the menu items are self-explanatory.

Overall, using a browser meant none of the annoying clunky and slowness that sometimes accompanies Java management applications.Monitors

ProactiveNet monitors require only about 31 KB of memory. And depending on what's being monitored and how frequently, our tests showed the agents consume only 1 percent to 2 percent of CPU resources.

The monitors collect data from systems, network devices, databases and applications, as well as Web, mail and security servers. Additionally, each category above has a range of monitors. In the systems group Host MIB, for example, information is supported as a source from which to gather data, but special collectors ship with the product--including those for Dell, Compaq, Windows NT/2000, BMC Patrol and Computer Associates' Unicenter--and these monitors require only configuration to work.

Complete SNMP data collection has been a part of ProactiveNet since version 1, and agents or monitors can now be distributed and used to collect non-SNMP performance data. Keynote and Mercury transactions have been added to the existing Web transactions.

ProactiveNet differs from most performance-monitoring applications in that it automatically sets performance thresholds based on historical trending. With other monitors, thresholds are set statically, requiring a deep understanding of what's considered normal for a given network, system or application; thus, it's not uncommon for the resulting alarms to be ignored once the thresholds are set.

The time it takes to set thresholds in this manner varies but, according to the vendor, six weeks is the norm. It took three weeks before the ProActivenet on our network had enough history to make sense of our thresholds. If the idea of fluctuating thresholds makes you nervous, you can set a static threshold as well.

ProactiveNet supports administrative access, as well as a couple of levels of user access. User access is assigned by the administrator and can be restricted to specific groups and/or views of devices, with or without read-only access. This doesn't allow for delegated access control, where access to a subset or group of devices can be further partitioned and authorized for access by an admin of that subgroup. It does, however, offer enough access within a single organization to cover most enterprise needs, because access levels are configurable.The system comes with three configurable user levels, one of which is read-only. The levels let users acknowledge alarms, view agent status and manage reports, as well as perform other tasks that can be combined and assigned by the administrator to provide the granularity of access deemed appropriate.

Finding the Source

The Diagnostics Wizard lets clients capture relevant data from a device to identify the source of a problem. The wizard comes with a suite of out-of-the-box diagnostics that have been expanded to include top 10 EJB (Enterprise Java Beans) for WebSphere, top 10 servlets for WebSphere, top 10 processes by CPU for AIX and Linux, a Logfile Dump mechanism, and ActiveSocket Dump and Execute (WebLogic).

ProactiveNet's new Domain Knowledge Capture links the user to a description of, and special instructions regarding, a problem on the network. I used this to create a link from an expected alarm that would occur during planned outages of servers and network devices. That way, whoever receives this alarm in the future will see that the service failures were identified properly, and will have instant access to any necessary corrections.

Overall, this release brings organizations a tool that, with a little tuning, can provide production monitoring of a network and everything that relies on it.Bruce Boardman is executive editor of Network Computing. Write to him at [email protected].

Post a comment or question on this story.