When Performance Monitor fails to get data it requests, or when it receives uninterpretable data, it displays a zero as the counter value for that request. The data request might have failed because of a malfunction or simply because the computer or application it was monitoring had stopped. Data considered uninterpretable includes negative times, negative performance values, or percentages greater than 100.
Performance Monitor does not alert the user when it gets bad data or no data. However, you can configure Performance Monitor to log these incidents to the Event Viewer application event log as errors and warnings.
Note
If Performance Monitor receives no data or bad data for the Processor: % Processor Time or the System: % Total Processor Time counters, it displays 100% as the counter value, not zero. Performance Monitor actually monitors the thread of the Idle process on each processor. It calculates the difference between 100% and the percentage of time the Idle threads ran. If a request for data on an Idle thread fails, Performance Monitor assumes it is zero and displays the difference, 100%, as the counter value.
Performance Monitor logs warnings and errors when it fails to retrieve data or receives bad data, and when connections to remote computers it is monitoring are lost and restored. Performance Monitor logging is either on or off; there are no intermediate or more detailed levels.
To log Performance Monitor errors and warnings to the Event Viewer application event log, use a Registry Editor, such as Regedt32.exe. Add the ReportEventsToEventLog value entry or change its value to 1, and then restart Performance Monitor.
Subtree | HKEY_CURRENT_USER |
Key | \Software\Microsoft\PerfMon |
Name | ReportEventsToEventLog |
Type | REG_DWORD |
Values | 0, 1 (0 is the default) |
After you have enabled Performance Monitor event logging and restarted the system, check the Event Viewer application event log routinely. You can use the Find or Filter Events options to display events with Source = PERFMON.
Many warnings and errors are attributable to normal and expected events such as processes or computers being stopped. Also, counters for threads are prone to uninterpretable values when the threads are stopped.
However, if Performance Monitor is logging negative time errors, there might be a problem with your HAL DLL. In the past, this has been encountered in some HALs for multiprocessor computers.
For more information on thread counters, see "Monitoring Threads and Processes."