Without a doubt, the Win32-based Service Reporter was the coolest, most functional out-of-the-box code we tested. We simply pulled the string to set up this standalone application, and the next thing we knew we were reading sophisticated usage reports via a Web browser. The setup routine finished its process, then it discovered our agents, created reports and published them in HTML. All the reports were preconfigured and sorted by Top CPU Busy, Top Disk, Top Network, Top Memory, Top Applications Used, Most Active Transaction, Transactions by System Availability and systems by OS grouping. The reports definitions were then set to run by default at 00:30 daily. All of this occurred just by running the setup.

We were even more impressed with the HP product when we compared the reporting and trending available from CA and Tivoli. Both required significantly more effort to yield any return. We also liked the fact that the reports in Service Reporter were easy to change with Crystal Reports templates. Making modifications and expansions was simply a matter of changing or adding the report definitions to the correct directory. Once the changes were added, the reports were automatically picked up on the next reporting pass and the modifications were reflected in the HTML reports. The online help includes a practical benchmark page. It describes HP's sizing reports, with straightforward hardware recommendations and results.

HP did not include its ITOperations (ITO) enterprise event management console in its pricing and took a hit in comparison to Boole and Babbage, CA and PLATINUM in regard to event processing. PerfView Monitor provides an event monitor but it lacks the sophisticated correlation that is available in ITO.

Computer Associates International Unicenter TNG 2.1
CA has a reputation for being a relentless competitor, and when it comes to performance management, it shows. In every category but price, CA's feature set is neck and neck with the competition. While some offerings--security and backup, for example--are available as separate products, you have to buy Unicenter TNG to get performance management. However, the price for Unicenter TNG is not so out of line considering that network and event management are included.

The Unicenter TNG 2.1 management console requires Windows NT. However, version 2.2 will add support for a Java console. This should make TNG more attractive to multiplatform sites.

Our biggest complaint with Unicenter TNG is that occasionally the degree of difficulty is over the top. Part of the problem is the contradictory design styles of each TNG application and the lack of consistency between GUIs. Even when the application was an Excel macro, we often wondered how something so seemingly simple could be so difficult to figure out.

Famous for its 3-D view of the networked world, Unicenter TNG is really a universe unto itself. Drilling into the managed system unearths many separate agents, including a simple ping daemon, systems agents, a real-time agent and, of course, a historical systems-performance agent. The 3-D view offers unified hierarchical views from a network perspective and contextual launching of the performance application. However, we soon found it was quicker to go directly to the performance application to find the layers we wanted to see.

TNG Historical Performance Agent's (HPA) configuration is both compact and flexible. The configurations are grouped into sections so that different statistical metrics can be collected at different rates throughout the day. Each section collects and sends the gathered data at variable rates to various overlapping or discrete TNG servers. This data, referred to as a "cube," is stored in profiles, which can be subscribed to by systems along OS lines or by any grouping of systems. The GUI organizes systems into an Explorer-like interface, making it easy to visualize and track the configuration. We found the cube approach much easier to use for auditing installed configurations than the profile approach used by Tivoli Decision Support.

Trend projection was a much more complex task with Unicenter TNG than with HP's suite. You use an Excel macro to set up the projections. Seeing this, we thought it would be a piece of cake. But it turned out to be incredibly difficult. We spent hours reading and experimenting and scratching our heads. We did finally get data out of our Solaris agent (NT and NetWare are supported but did not work in our tests), and uncovered trend algorithms for projecting future usage.