Sams Teach Yourself MCSE Windows NT Server 4 in 14 Days
(Publisher: Macmillan Computer Publishing)
Author(s): David Schaer, et al
ISBN: 0672311283
Publication Date: 12/15/97

Because the log file can accumulate over a period of hours or even days, it is the best method to use both when you are establishing a baseline and for routine monitoring. You also can export data from log files either to a spreadsheet or database program for further evaluation.

You can insert bookmarks into the log to identify major events. Using bookmarks and alerts enables you to analyze a large log file with considerable speed.

Figure 10.2.  In the Log view, the Performance Monitor displays the log filename, log interval, and log file status.

10.2.3. The Alert View

The Alert view enables you to monitor multiple counters or computers with minimal overhead. You must set a threshold value for each counter, as shown in Figure 10.3. For example, if the amount of used pagefile space was to exceed a certain threshold, the system could log the event automatically, notify an administrator, or execute an application. You can configure more than one alert at one time; however, this is processor-intensive, and you should do this only when necessary.

10.2.4. The Report View

As the name implies, the Report view, as shown in Figure 10.4, presents the selected data in a report format. Objects display in the order you select them. The Report view gives you a quick picture of the available data to assist you in selecting the objects you want to monitor. There is not an option in Performance Monitor to print the report, although you can export the data into another application, such as Excel or Access, to generate and print reports and graphs.

To reduce the amount of real estate Performance Monitor occupies, remove the title bar, menu bar, and status bar.

Figure 10.3.  An Alert set with a threshold > 80 %Processor Time.

Figure 10.4.  A Report view displays %Processor Time.

10.3. Selecting the Key Elements to Monitor

Knowing which are the key elements to select and monitor is essential to the proper evaluation of system performance. It is necessary for you to understand that although you can track individual objects, the first object you identify as performing poorly is not always at fault.

As you examine each of the following objects, you must consider how each object affects the performance of the others. For example, an abnormally high processor utilization average is often the result of insufficient memory.

When you select which counter to monitor, remember that the default counter usually provides the most significant information about that object.

Each object you track is matched by a list of corresponding counters and instances. For example, in Figure 10.5 the selected object is the Processor; because this is a single-processor system, there is only one instance, 0. The counters are the specific areas of the object you might track; in this case, the counter is %Processor Time.

Figure 10.5.  Select counters in the Add to Chart dialog box.

When selecting objects, use the Explain button in the Add to Chart dialog box to display a short description of the purpose of each counter.

10.3.1. The Processor Object

The Processor object,as shown in Figure 10.5, displays the percentage of time in which the processor executes the active threads from both user and system processes. Before you decide that a processor upgrade is required, look for a lack of memory in the system. Excessive paging can increase significantly the work that both the processor and disk subsystem require. If you are monitoring a multiple-processor system, be sure to monitor each processor.

Don’t concern yourself overtly with sporadically high readings, or spikes. The key to watch for here is whether the percentage processor utilization remains consistently above 80%.

10.3.2. The System Object

The System object and its related counters reflect the collective performance of objects. Two common counters are set with the System object.

%Total Processor Time is useful in a multiprocessor system, where there would be a separate instance for each processor in the system tracked by the Processor object. In such a system, you can determine an average percentage of processor utilization across all processors by tracking the System object and the %Total Processor Time counter, as shown in Figure 10.6.

Figure 10.6.  Notice that no instances are listed because this counter is reported as a total amount.

As shown in Figure 10.7, you can track the total interrupts per second over a period of time in the Log view to create a baseline. An abnormal increase in the Total Interrupts/second could indicate a potential component failure.

Figure 10.7.  Again notice that no instances are listed because this counter is reported as a total.

10.3.3. The Memory Object

The Memory object includes a series of counters that provides an insight into how the system is utilizing its memory.

The Pages/second counter of the Memory object is the most critical counter to track.

Pages/second, as shown in Figure 10.8, is the most critical memory counter. This number will be high when a significant amount of the data you request cannot be located in physical RAM.

Figure 10.8.  The Add to Chart dialog box shows the Pages/sec object selected.

A lack of memory causes a multitude of inadequacies. Given a choice between a faster processor and more RAM, choose additional RAM.