TDS is a bright spot. This suite of Cognos PowerPlay, a multidimensional OLAP (online analytical processing) product, and Crystal Reports has just begun shipping. Tivoli is gluing them together with predefined Crystal Reports definitions and predefined cubes for Cognos. The cubes define data and its relationship to information from other Tivoli Framework applications. Tivoli provided us with an event cube that let us look at Framework events from Distributed Monitor by date, machine and hardware/operating system. This is similar to HP's Service Reporter, but it has added potential to relate help-desk calls, server-utilization spikes, and network-node availability back to service-level agreements.

Tivoli surprised us on the pricing front, and the uncharacteristically low cost kept its solution in our competition. Tivoli's policy is to quote retail prices only in order to protect its distribution channel, and this policy usually places the company's products in the budgetary stratosphere. But this time the base framework of common services--Distributed Monitor and TDS--came in at a price way below the other products. And it offered the best warranty and competitive service options.

Although we did not test the Tivoli Enterprise Console--it boosts the price tag and Tivoli decided not to include it for this comparison--it is a significant part of the solution. Without the console, the distributed monitoring had passable eventing, status-sensitive, color-coded icons and pop alerts.

Tivoli Enterprise is all about organization. Distributed Monitor is no different. It fits tightly into the Tivoli Policy Regions, and every aspect from configuration to event creation can be designated. This seems compartmentalized at first, but as we used it more, we liked the structure's inherent control and audit capabilities.

Agent configuration must be controlled with this approach, as it is easy to lose track of what is running where. Tivoli suggests that a single configuration profile be specified for all agents within a policy region. Of course, we disregarded this advice, thinking that it would be best to divide statistical metric collection along hardware/operating system lines. As it turned out, the framework's common services can delineate which metrics are applicable to which hardware/operating system combination. Having all statistical collection definitions in a single profile reduces the number of configurations that must be maintained.

However, in CA's case the configuration tool is easy to use and fairly flat in its display. By comparison, Tivoli's configuration tool requires a number of text-based edit screens to complete. The summary display of configured metrics was very good, but we found that when modifying metrics, the Tivoli Desktop would occasionally shut down completely upon saving the edit. Tivoli says it is working on a solution.

The Tivoli Inspector is a very weak Web-based display of statistical metrics. We were able to display some basic metrics, but gave up after repeated failures to consistently display correctly configured statistics running in real time.

Boole and Babbage Command/Post
Flexible agents coupled with the ability to handle events from any platform are the primary strengths of this veteran performance-driven event engine. Boole and Babbage makes no excuses for its lack of diagnostic prowess. Instead, the vendor proudly claims the event correlation crown. We'll grant it that much, but event management is only a slice of what you need for systems performance management.

Like the other solutions in this review, Command/Post is comprised of many applications. Boole and Babbage's proprietary performance agents are called PowerModule agents, and the primary interface for configuring them is Command/Post Explorer. Running on Unix or NT, Command/Post Explorer can be used to control and display agent configuration, events and operator assignments. Command/ Post--a separate product--is a manager of management consoles. It not only collects PowerModule events and alerts, but also will collect events from any SNMP device, BMC Patrol Agents, PLATINUM's Network Monitor, and Tivoli, CA and HP event consoles.

The installation of our test Solaris7, NT and NetWare servers went very smoothly and setting up the agents was as easy as deploying the agents in the HP solution. Getting the agents up and running was simply a matter of making some fairly easy text-based edits. As with the HP product, we missed the auditing function that a GUI configuration can provide. However, we had no trouble understanding the parameters for setting up the agent.

One of the strongest features of the PowerModules is the ability to track an impressive number of statistics right out of the box. In addition, the PowerModules can initiate local action without relying on a centralized event correlation engine. We set up metrics that watched for delta changes of statistical metrics. When agent changes exceeded our thresholds, alarms were sent to our Command/Post Explorer and Consoles.

The PowerModules provide a healthy list of canned capabilities. They are built around housekeeping-type administrative chores, such as monitoring and deleting core files, checking password files for illegal entries, CPU performance, disk utilization, network errors, swap and Super User failures. We found the manuals and online help for the PowerModule options clear and concise.

Rather than looking at the metric in a table or graph, Command/Post focuses on applying thresholds and events to their scans for operations notification. For example, a scan might trigger the erasure of a core dump when low on disk, or a check of a CPU reporting a warning when greater than 80 percent in five minutes. When you consider this type of capability, in addition to the basic checking of average, minimum and maximum values as part of the base metrics, it's easy to see the immediate payback. The PowerModules are also easy to deploy.

Command/Post has configuration profile grouping called packages. These are groupings to be collected and scripts to be run, along with thresholds and local actions. They're similar to CA's cubes in that the packages offer an efficient, consolidated approach for managing and distributing agent configurations. Command/Post is impressive when it comes to event management, but the lack of diagnostics makes it too focused for performance management.

Systems Wrap-Up While HP took our Editor's Choice award, the two runners-up, CA and PLATINUM, were only a hair behind. Each of the three top products will do an admirable job for sites that need strong systems performance management from eventing to diagnostics to trend analysis. Both Tivoli and Boole and Babbage have strong backgrounds in event-oriented operations. But while Tivoli is leveraging its agent technology to provide strong trend analysis, Command/Post is unlikely to move beyond its event exception heritage any time soon.

Send your comments on this article to Bruce Boardman at bboardman@nwc.com.