Like the other vendors, HP bundles multiple applications and products to create a total systems management solution. HP has taken a page from PLATINUM's book and created integration where it counts. Unlike PLATINUM, CA and Tivoli, HP has not overloaded its solution with a common framework or platform.

Besides the MeasureWare Agents (MWA), HP provides GlancePlus for real-time performance measurement, PerfView Monitor for historical views, PerfView Planner for trending and projection of baselines, and Service Reporter for multidimensional analysis of diverse but related statistics. They are all in the OpenView enterprise systems management family.

The capability that sets the HP solution apart is its out-of-the box collection of statistical metrics. Once the various consoles were all working in concert, it became apparent that HP's reporting put it ahead of the other solutions. Just like that, we were able to get reports on multiple complex systems and collect multiple statistics with no setup problems, no learning curve and no hassle.

Regarding diagnostic tools, our testing uncovered comparable offerings in CA's Unicenter TNG and stronger analytical potential in Tivoli's TDS. However, we hit a lot of sour notes trying to access the meaning of the performance data. In contrast, HP has taken the time to make sure that its products are useful and not just a checklist of the "right" features.

Anchoring performance is the MWA, HP's solid--not flashy--agent technology. The agents read all statistics from the Kernel Interface on every collection pass. The beauty of this method is that no matter how busy the system became or how many metrics we collected, the overhead remained constant. We used HP's annotated configuration examples along with all of the HP monitoring tools to monitor this agent load.

Configuring the MWA is somewhat difficult because HP provides only a command-line interface to the configuration file. All of the other products, particularly those from CA and PLATINUM, use a very functional GUI. But even without a GUI, the annotated files and the sample configurations made for an easy setup and HP does provide syntax checking.

However, HP's solution has an Achilles' heel in this area: Its inability to track configurations and subscriptions. The other products we tested, particularly CA's Unicenter TNG, do a good job with this task.

PerfView Monitor provides an excellent historic view of systems performance. The simple line graph interface is identical on both NT and Unix, and belies the need for sophisticated drills to get more detailed statistics. We not only looked back in time at groups of statistics on multiple systems, but also looked more closely at the raw statistics for any given period. For example, we drilled back in time, went from daily samplings to monthly to a sampling every five minutes. We then dipped into the applications running at the time, the network traffic, the processes, disk activity and the specific numbers that made up the data points of the graph. The graph served as a high-level guide from which we could drill down.

PerfView Planner, which forecasts future trends, also has a simple interface that permits sophisticated trending. From a single configuration screen, we chose what statistics to project, how much history we would use as a base and the method for our projection algorithm (linear, exponential, or s-curve, smoothed or as a business unit). We also chose the confidence level at which we wanted our guesses to be made. The help files did a good job explaining what projection method to use. We were especially intrigued by the business unit method. This provides for a collection of external statistics to be compared and projected against actual systems performance. For example, we could forecast the number of known sales transactions against actual CPU, memory, disk and network activity.